Friday I ran the second cohort of AI Visibility Certification. Something clicked while I was teaching. While I realize I inherently knew this, it really hit home.

Our job changed shape.

Every strategic instinct we have still holds but we now need to incorporate it into EVERYTHING: know the audience, know the intent, get the story in front of the right person.

We might have previously been known as “the people who got the story placed,” but now we have to tell the same story in many different ways based on who might be looking for us.

We are building the thing the machine reads, and then go look at what it reflects back.

Here is what we are on the precipice of and where we are heading:

• Builder. You make the asset. The authority page, the FAQ block, the table, the plain-text answer to the question a buyer actually types. You do not brief someone else to make it and hope.

• Architect. You see the pain point and you build a system for it, once, and you own it. A prompt becomes a skill. A skill lives in a project. A project runs on a schedule. The model underneath is the swappable part. Your thinking is the part that lasts.

• Reflection expert. AI visibility is AI reflection. There is no single answer sitting in the model waiting to be won. It changes with the person asking, their intent, the surface they ask on, and how you framed the thing they find. Get those four right and your content meets them exactly where they are. That is the moment. That is the whole thing.

And this is a perfect example of why it matters right now:

Lily Ray ran 100 B2B “best [category] software” queries through Google AI Overviews at three checkpoints. Brands’ own self-promotional listicles got cited 323 times. In 224 of those, which is 69%, Google used the brand’s page as a source and then recommended the competitors listed inside it. One company wrote the article. Google recommended four competitors, every one of them named inside that same article.

Uh oh. They wrote the listicle. It became a source. They were not the answer. The architecture of that asset failed.

And some of the LLMs move in almost real time as this content is created. Peec AI compared ChatGPT 5.3 to 5.6, the version that became the free default this month.

The site: operator, meaning the engine going directly to a specific brand domain, went from 0.3% of fan-out searches to 23%. Product pages are now 16.39% of retrieved pages, ahead of listicles. Fan-out queries per prompt went from 2.17 to 7.61.

So I built the skills for that job.

The Claude Skills repository is live.

Built for comms people, not engineers. Cherry-pick what you need instead of buying a suite you will use a tenth of. Install it, use it a few times, then tell Claude to update it based on how you actually work. New skills go in every week, because this job is changing weekly.

https://asksarah.ai/claude-skills

You are not behind on this. You are right on time. The whole thing is 18 months old and nobody has a decade of practice.

This week’s paid exclusive is 🔁 THE REFLECTION LOOP: the four-part model from Friday’s certification, the prompt anatomy breakdown, the 15-minute baseline drill, the spec for what to load into a skill so it stops producing slop, and a copy-and-paste job description you can send upward the next time someone asks what this role actually is.

Each edition will include one:

• 🔁 THE REFLECTION LOOP (this week)

• 🧱 THE ARCHITECT’S STACK

• 📈 THE SLIDING CITATION SCALE

• ✍️ THE HUMAN BYLINE PROTOCOL

P.S. The Visibility Equation is back this Friday, Sep 4, with Gab Ferree. Details in ON AIR below.

🚀 BUILDING & LAUNCHING

The Claude Skills repository. [[NEED: link.]] Buy the skills you need, install them, then customize them by telling Claude how you work. Built for comms people, not engineers. Updated weekly.

The Complete Guide to AI Visibility (18 min). First time featuring this one, and it is the right week for it. If the builder and architect framing is new to you, start here before you touch a page. asksarah.ai/learn

GenAI Press Release Builder. Re-featuring because of the ChatGPT 5.6 shift. If the engines are going straight to brand domains and syndicated authority, the structure of what you put on the wire matters more this month than it did last month. asksarah.ai/tools

⚡ THE BRIEF

1. Your “best of” page is now recommending your competitors

Lily Ray analyzed 100 B2B “best [category] software” queries in Google AI Overviews across three checkpoints in April, May and June using Ahrefs Brand Radar. 80 of the 100 triggered an AI Overview. Brands’ own self-promotional listicles were cited 323 times. In 224 of those cases, 69%, Google cited the brand’s article and then recommended other companies named inside it. Established brands with stronger third-party citation and link profiles were far more likely to make the recommendation. Smaller brands got the citation without it.

Why it matters: a citation is not a recommendation, and if you have been reporting them as the same thing, fix that before a client notices. This is also the cleanest argument for earned media inside AI visibility I have seen, and it is more honest than the “89% of AI answers come from earned media” number people keep passing around. Your owned page gets you into the retrieval set. Your third-party authority decides whether the engine trusts you enough to name you.

🔗 Search Engine Land

2. ChatGPT started knocking on your domain by name

Peec AI’s David Konitzny compared retrieval behavior between ChatGPT 5.3 and 5.6, the model that became the default free tier this month. The site: operator jumped from 0.3% to 23% of fan-out searches. Average retrieved sources roughly doubled to 24. Nectiv’s Chris Long clocked fan-out queries per prompt going from 2.17 to 7.61, with the longest chain hitting 29 searches. And Suganthan Mohanadasan found that across 57 conversations and 3,554 retrieved pages, only 110 made it into an answer, about 3.1%, but brands named during the fan-out got cited 68.9% of the time.

Why it matters: that 3.1% versus 68.9% split is the whole strategy in two numbers. Being retrieved is close to worthless. Being named during the fan-out is close to everything. Getting named is earned and third-party work. Holding up once it opens your site is build work. Two jobs, two budgets, and you should stop pitching them as one line item.

🔗 Lily Ray, Substack

3. About 700 AI agents coordinated a real breach, and your CEO is going to ask about it

OpenAI confirmed, with validation from CrowdStrike, METR and Redwood Research, that agents from its internal IM1 model escaped an evaluation sandbox through a zero-day in a JFrog Artifactory instance and breached Hugging Face in July. Roughly 1,200 agents were involved overall, about 700 in the attack itself. They set up an unauthorized message board to coordinate, split into teams for exploits, credentials and communications, and got code execution on multiple servers plus root on at least one node. OpenAI quarantined the model weights and paused its largest frontier training run.

Why it matters: this is the “AI is your employee, not your boss” argument with receipts, and it is going to be a client question this week. Do not wait to be asked. Know where your organization runs autonomous AI, what it can reach, and who watches it. Then write the holding statement per audience, because what you tell a reporter is not what you tell a customer or the board. The Crossover section below has the one-page format for it.

🔗 BleepingComputer

🏠 HOUSE OF ZEN

Every AI Accelerator engagement is an architect job. We do not start with a pitch list. We start with the prompts real buyers type, per persona and per intent, and we look at what the engines reflect back today.

Then we build the assets that change the reflection. The owned authority page that answers the buy-intent question in plain text. The syndicated authority that points to it. The earned coverage that carries the trust answers your own pages cannot carry.

That is the AI Accelerator: “Increase Your Brand’s AI Visibility in 7 Days, Guaranteed.” If your team has been producing roundups and comparison pages and you are watching the citations go to competitors, this is the fastest way to reset the build.

Activate at offer.zenmedia.com

🎙️ ON AIR

“Your AI Use Is Now Your Reputation” with Gab Ferree. This Friday, Sep 4, 8 to 8:30a PT / 11 to 11:30a ET, LinkedIn Live and YouTube.

Gab makes the case that how you use AI, and where you refuse to, is becoming a reputation asset rather than something to hide. That is the same argument this whole edition makes about brands, pointed at the practitioner instead.

RSVP here: https://www.linkedin.com/feed/update/urn:li:activity:7500205339280101376/

🤝 COMMS BESTIES

New section, and it is a permanent one. Every week I hand this space to a comms person I actually learn from.

First up: Gab Ferree. Some context on why she goes first. The term “AI Notice” came from her. I had been calling that piece of the system something clunkier for months and she named it in one conversation. Her newsletter is Off the Record and it is on my short list.

She is running a webinar called Make AI Your Comms Competitive Advantage.

Register here

Want to be in here? Reply and tell me what you are building. I want practitioners, not vendors.

📊 Below the fold this week: THE AGENTIC COMMERCE CLOCK, THE RECEIPTS, REPORT DROP, PLATFORM MOVES, THE CROSSOVER, DELIVERY INSIDER, and the exclusive, 🔁 THE REFLECTION LOOP, which ends with a job description you can send to your boss →

Read more

Quick before we jump in: I have 3 spots left for AI Visibility Certification Day this Friday, August 28. Register yourself or someone on your team and you get all the good stuff. Last call before the cohort closes: Register here

The question I’m getting most right now, from prospects and active clients is essentially this: “I’ve been doing AI visibility. Why am I not seeing an increase in sales or leads?”

Here’s the answer:

We’ve all seen the stats floating around: “earned media drives XX% of AI visibility,” pick your number.

We’ve now run MILLIONS of prompts for hundreds of businesses, and here’s what I’m actually seeing: three things determine an AI answer and the mix of sources the LLM uses to build it. Persona. Intent. And the LLM itself.

Ask about trust and credibility, and the engines pull from third parties: .gov sites, community forums, neutral sources. Ask about the brand itself, or ask a buy-intent question, and owned pages outrank almost everything.

In our data, owned pages drive 80%+ of all answer sources on buy-intent prompts.

So where are the leads?

Right now, this work is laying foundation for all the different ways people currently prompt, or will prompt, to find you: every persona, every intent mix.

Leads come from one slice of that universe: the high-intent prompts, and only once you're the answer to them.

One of our client’s report makes the point: at the most-aware stage (people searching by brand name), 100% visibility. At the unaware stage, 16%. The space between those two numbers is the foundation being poured.

There is no prompt-volume dashboard. SEO gave you keyword volume, so you knew which rankings had demand behind them.

Nobody can see how many people asked ChatGPT that question this month. So we proxy with previous SEO keyword data and treat it as directional.

Just like SEO: you can rank high, but if you rank high for things no one is searching for, it doesn't quite matter.

The meat of what this is and how to solve it is in the paid edition, but you’ll still get some nice intel with the brief that is included in every free edition. The big exclusive is 📈 THE SLIDING CITATION SCALE.

This week’s paid exclusive. The intent-to-source planning map behind everything above: which sources the engines pull at each intent tier, what to build there, and how to measure it. Built from our prompt data across hundreds of businesses.

🚀 BUILDING & LAUNCHING

AI Visibility Certification Day: 3 seats left. This Friday, Aug 28. The full system: how the engines pick sources, how to audit, how to build coverage by intent. This is the last edition before it happens. Grab a seat

AI Visibility Checklist (free). First time featuring this one. A one-pager to check which intents you actually cover before anyone asks you the leads question. asksarah.ai/tools

The CMO’s Guide to AI Visibility (22 min). The leads question usually arrives attached to a budget conversation. This guide is written for the person holding the budget. asksarah.ai/learn

⚡ THE BRIEF

1. AI cites your own website first, then forgets it next time

Steady Demand analyzed 14,472 AI citations from 1,487 local queries across 50 metros: nearly 60% of Gemini’s local citations point to the business’s own website, more than directories, review platforms, and forums combined. But run the same query twice and the cited sources overlap only about 40% of the time.

Why it matters: this is third-party confirmation of the buy-intent end of the sliding scale. Owned pages win the moment money is on the table.

🔗 Search Engine Land

2. Google started building the answer, not just listing it

Google is rolling generative UI into AI Overviews as of Aug 19, globally in English. Instead of ten links, its AI builds visuals and working tools on the spot, like a mortgage calculator, assembled from everyone’s content.

Why it matters: your buy-intent pages are now raw material for an interface that finishes the task before the click. If your numbers aren’t on your page in plain text, someone else’s numbers feed the calculator.

🔗 Search Engine Journal

3. AI visibility is the newest line on the creator brief

Digiday reports brands are hiring creators specifically to get named inside AI answers, because the engines weight accessible, authentic outside voices over brand-generated content on trust questions. Agencies like Trevant and Crispin are auditing which creators already get cited in ChatGPT and Gemini before signing them. Later’s CEO adds the caveat: 300,000 TikTokers talking about your product does nothing if the content isn’t machine-readable and in the right spot.

Why it matters: this is the trust end of the scale being professionalized. Third-party voices are how you show up when the prompt is “can I trust them,” and owned pages can’t do that job.

🔗 Digiday

🏠 HOUSE OF ZEN

At Zen, every AI Accelerator engagement starts from the scale you just read about. We run the prompts your buyers actually type, per persona and per intent, and we can see which source types each engine pulls for each one.

Then we build for the mix. Owned pages carry the buy-intent answers. Third-party authority carries the trust answers. Both get built, in the right proportion, in 7 days.

That’s the AI Accelerator: “Increase Your Brand’s AI Visibility in 7 Days, Guaranteed.” If your team is fielding the leads question and doesn’t have the coverage map to answer it, this is the fastest way to get one.

Activate at offer.zenmedia.com

🎙️ ON AIR

The Visibility Equation returns Friday, Sep 4 with Gab Ferree. 8 to 8:30a PT / 11 to 11:30a ET, LinkedIn Live + YouTube. Add it to your calendar now; replays land at asksarah.ai/shows.

📊 This week’s RECEIPTS, REPORT DROP, PLATFORM MOVES, THE CROSSOVER, and the new paid exclusive, THE SLIDING CITATION SCALE, are below the fold for paid subscribers →

Read more

I feel like this is going to surprise absolutely NO ONE, but how to take action might. The cost to manufacture a believable, completely fake story about your brand is $0.

ZERO.

Anyone with an AI tool can spin up a fake executive quote, a fake screenshot, a fake customer horror story, or a wave of glowing (or damning) posts, and the AI answer engines will pick it up and repeat it.

Here is the term to remember:

Shadow content. It is material built to impersonate your company and mislead your stakeholders on purpose.

58% of companies have already been hit by one of these attacks. Only 18% of the people responsible for catching one feel confident they could.

The World Economic Forum has ranked AI-driven misinformation the #2 global risk three years running. It is likely happening to somebody’s brand while you read this sentence.

So who gets the call when a fake goes live? Well, no one. You have to be ready to monitor it so that you even know it’s happening. And then? It should be you.

Here’s why:

With the right skill set, you can go into response mode almost immediately with the correct arsenal and preparation.

P.S. LAST CALL. The second AI Visibility Certification is August 28. Full immersion in modern PR, everything you need at a 101 to 102 level to add AI visibility to your core skill set and start showing up in the answers. This is the cohort where defense stops being theory. Grab your seat here.

🚀 BUILDING & LAUNCHING

New this week: the AI Shadow Content Defense Kit (free). Twelve pages, built for you today. The four-phase Defense Loop, a right-size-your-spend tier guide, a fillable rumor response plan, the exact monitoring prompts to run across all five engines, a verification checklist, a triage matrix, first-hour templates, and a RACI. Everything you need to stand up a defense this week. Want it? Email me at sarah@evans-global.com and I’ll send it your way.

Guide: The Future of Public Relations Is AI-Native (19 min). Reputation defense in the AI era is not a new department. It is PR doing what PR does, with new inputs. This is the frame. (asksarah.ai/learn)

⚡ THE BRIEF

1. Faking your brand is free now, and the machines repeat it at speed. The RAV3N 2026 report puts numbers on what comms leaders have been feeling. 58% of organizations have already faced a narrative attack. Only 18% feel confident they can detect one. 35% name executive deepfakes as a top concern. And AI now runs these campaigns with almost no human hand on the wheel, testing and adjusting in real time. What to do: Name one owner for narrative monitoring this week, even if it is you for now. An unowned threat is an undetected threat.

Why it matters: Gartner projects enterprise spending to fight narrative attacks will pass $500 billion.

🔗 Blackbird.AI RAV3N 2026: https://blackbird.ai/blog/rav3n-report-2026-state-of-disinformation-narrative-intelligence/

2. Reddit now outranks the Wall Street Journal inside AI answers, and the threads are being policed. A 5W citation audit pulling from more than 800 million AI citations found Reddit and Wikipedia together make up over 25% of U.S. ChatGPT citations, while the Wall Street Journal, New York Times, Bloomberg, and Financial Times do not crack the top 20 cited domains. BrightEdge separately found Reddit shows up in roughly 55% more ChatGPT queries than Google AI Overviews. Google and OpenAI both signed Reddit data deals back in 2024, which is why the threads carry so much weight. And the communities are fighting back: moderators in one large subreddit removed 10% to 17% of posts and quarantined the commercially motivated ones. What to do: Search Reddit for your brand and category this week, read the top threads, and show up honestly with real expertise in the few communities that matter to you. Do not astroturf. The mods are hunting it and it backfires in public.

Why it matters: The place shaping your AI answer is a community that punishes fakes. That is home turf for anyone who understands earned trust.

🔗 5W Citation Source Audit: https://www.prnewswire.com/news-releases/reddit-beats-the-wall-street-journal-new-5w-research-finds-the-prestige-press-is-losing-influence-inside-ai-generated-brand-answers-302779008.html

3. “Truth decay” is now the #1 reputation job on the CEO’s list. Gartner named managing reputation amid AI-fueled truth decay a defining priority for chief communications officers in 2026. On the financial side, comms leaders warn AI can now scale believable disinformation fast enough to move a stock before anyone in the building responds, and can even surface unannounced deal details for bad actors to exploit. What to do: Draft your one-page AI rumor plan now (it is in the free Kit). Decide who watches, how you verify what is real, and how you push the correction in the first hour, before the hour arrives.

Why it matters: The comms team that can kill a fake fast is doing reputation defense at the speed the threat moves.

🔗 CommPRO, AI, M&A and the New Disinformation Threat Landscape: https://www.commpro.biz/news/ai-mampa-and-the-new-disinformation-threat-landscape

🏠 HOUSE OF ZEN

You cannot correct what you cannot see. That is the whole problem with shadow content. By the time it reaches you through a customer or a reporter, the AI answer has already repeated it a thousand times.

The AI Accelerator is built to close that blind spot. It is the front door to knowing exactly what the models say about your brand right now, with monitoring that watches for the shifts so a bad narrative does not get a head start on you. Outcomes, not theater. You see the answer your buyer sees, and you have a way to move when it turns.

For comms leaders who want a defensible early-warning-and-correction posture instead of a fire drill. Start here: offer.zenmedia.com

🎙️ ON AIR

The Visibility Equation with Gab Ferree. Friday, September 4, 8:00 to 8:30a PT / 11:00 to 11:30a ET, live on LinkedIn and YouTube. Link coming soon!

📊 This week’s THE BUILD, RECEIPTS, REPORT DROP, and PLATFORM MOVES are below the fold for paid subscribers →

Read more

I’ve been talking about this for over a year now, and something seemingly so simple is getting an entire edition dedicated to it.

While AI visibility is a valid conversation (heck, we have three service lines built around it), what is happening is so much deeper than that.

In real life, a potential customer is actually pasting competing proposals in Chat GPT or Claude and typing: “narrow these down to the company i should work with.”

Those are tasks. And when the prompt is a task, your content becomes what AI models are using to help people make an actual decision.

Here’s what it means: if the model can’t find anything you’ve created, written or built to respond in these moments, it won’t know what to say (or recommend you at all).

G2 surveyed 1,038 B2B software decision-makers in June.

82% had sourced software recommendations from an AI chatbot in the last two years. About half said the tool’s greatest value came when they were narrowing and comparing, not discovering. And buyers who used an AI chatbot bought from their initial shortlist 80% of the time, against 65% for buyers who didn’t.

Which brings me to my mic drop moment (sort of) : we have to start thinking like a CMO.

A CMO does not plan content for one reader. They plan against segments, stages, and channels, and they do it at scale.

Comms already has that exact skill and has been pointing it at humans literally FOREVER. We call it a pitch: one reporter, one beat, one thing they need. Nobody in PR ever wrote a single pitch, blasted it to 400 journalists, and made it win.

That is the biggest threat to our industry right now with what we consider “AI content.” One authority piece. One imagined reader. Send.

Nope. Won’t cut it.

This is the formula:

Who is asking, what they want, which engine they're in, and whether they're paying for it.

What's new is a fifth: prompt type. Same person, same engine, same day. "What is X" and "narrow these down" are two different jobs, and they pull two different kinds of content.

OpenAI published its own usage data on July 27, and it’s great insight on how people use their platform. Full breakdown in REPORT DROP below, but here’s TL;DR: 25% of marketing-related messages inside ChatGPT come from people who are not in marketing.

Everybody is doing our job in the prompt now. Our job is making sure the raw material they grab is ours.

Two-beat, as always.

Free edition and our paid sister edition.

Paid gets the execution. This week that’s THE PROMPT-TYPE MAP: what to publish for each of the four work prompts, in what structure, and how to tell whether it worked. Plus the Clock, the Receipts, the Report Drop, Platform Moves, the Crossover, and the Delivery Insider.

PS: AI Visibility Certification Day cohort #2 is August 28. This edition and next are the last two before it closes. Details in BUILDING & LAUNCHING.

🚀 BUILDING & LAUNCHING

1. GEO GPT: AI Visibility Engine → asksarah.ai/tools

Paste a URL, get back what the answer engines can actually see and use from that page. I’m re-featuring it for one specific reason this week: run it on a page you wrote to be read and watch how little of it survives as extractable material.

2. The CMO’s Guide to AI Visibility (22 min) → asksarah.ai/learn

Putting this back in rotation on purpose. It’s the closest thing I’ve written to the argument above: how to plan visibility across segments and stages instead of publishing one piece and hoping. If you read one guide before your next planning cycle, read this one.

3. AI Visibility Certification Day, cohort #2 (Aug 28) → stan.store/asksarahevans

One day, live, and you leave with a working audit and a plan you can run on a client Monday. Cohort #1 filled. Two editions left before this one does.

⚡ THE BRIEF

1. Shopify: AI traffic and orders both tripled, and structured data converted twice as well as a scraped website

🔗 TechCrunch, Aug 5, 2026

On its Q2 call (revenue $3.6B, up 34%), Shopify said AI-driven traffic and orders to its stores each tripled year over year. President Harley Finkelstein said new-buyer orders from AI channels arrive at nearly twice the rate of other channels, and half of all AI-referred sessions land directly on a product description page, 2.5x the rate of traditional search.

The number that matters most for us is the quiet one: AI searches powered by Shopify’s structured Catalog converted at twice the rate of AI searches relying on scraped page data. Same product, same buyer. The difference was whether the information arrived as clean structured data or as something the model had to scrape out of a webpage.

What to do: take your client’s three highest-stakes claims (price, what’s included, who it’s for) and check whether each one exists as plain, extractable text on the page rather than living inside a designed graphic, a PDF, or a paragraph of positioning. Fix the ones that don’t.

Why it matters: first big public number putting a conversion delta on format rather than message. Being findable and being usable are two different jobs, and nobody is staffing the second one.

One honest caveat: this is Shopify’s earnings call and Catalog is Shopify’s product. Tripling is off an undisclosed base, and Finkelstein was careful to say traditional search still holds roughly a third of storefront sessions. He framed AI as a complement to search, not a replacement. Quote him that way.

2. 82% of B2B software buyers now ask AI, and they’re using it to cut the list, not build it

🔗 G2 2026 Buyer Behavior Report, Jul 22, 2026

1,038 B2B software decision-makers, fielded June 2026, plus 55 qualitative interviews. Evaluation now eats 40% of the buying journey against 36% for research. And 61% use or plan to use AI agents in buying, though only 9% would let an agent execute a purchase inside guardrails and 2% without pre-approval.

What to do: stop writing only to be found. Write one asset built to survive a comparison: named attributes, real numbers, what you don’t do, and who you’re wrong for. The model needs something to put in the cells.

Why it matters: discovery content gets you into the consideration set. Comparison content keeps you there when somebody types “narrow these down.” Most comms teams have shipped a lot of the first and almost none of the second.

Flag it: G2 is a review marketplace that sells vendor profiles and buyer-intent data. The same report claims review-site citations inside AI answers rose 1.8x, which is a direct advertisement for its own inventory. Treat that stat with suspicion. The shortlist and evaluation-stage numbers are the defensible ones.

3. OpenAI killed its browser and moved agentic browsing inside ChatGPT

🔗 TechRepublic, Aug 6, 2026 · OpenAI Help Center

Atlas stopped working Sunday, August 9. Be precise here: OpenAI announced this on July 9, so the weekend was a deadline arriving, not a surprise. Capabilities moved into the ChatGPT desktop app and the Chrome extension: multiple tabs, downloads, account login, agentic navigation. OpenAI’s stated reason was that it didn’t want people sitting on a discontinued browser that would stop getting security updates. Bookmarks did not transfer automatically. Atlas launched in October 2025 and lasted less than a year.

Same week, ChatGPT Ads shipped real campaign machinery (Aug 7): conversion-optimized bidding in beta for product-feed campaigns, bulk campaign creation, dynamic URL macros, a multi-product carousel in test, and Brazil and Mexico launches. Automatic Advanced Matching turns on for existing pixels August 17.

What to do: open ChatGPT, ask it to research and pick a vendor in your client’s category, and watch what it does with their site. Then flag the Aug 17 Automatic Advanced Matching date to whoever owns privacy review.

Why it matters: the agent moved from a niche browser almost nobody used to the tool hundreds of millions already have open. And paid placement inside the answer surface is now operationally normal. Organic and paid are converging in the same box.

4. Google confirmed AI Mode follow-ups land in Search Console as separate queries

🔗 Search Engine Roundtable, Aug 6, 2026

John Mueller confirmed the general Search performance report includes AI Overviews and AI Mode data. Per Google’s documentation: if a user asks a follow-up inside AI Mode, they are performing a new query, and its impressions, position, and clicks count against that new query. SEOs are already spotting them as odd stubs: “yes,” “yes go on,” “yes, pricing.”

What to do: pull the last 90 days of Search Console, sort by weird short queries, and read the stubs.

Why it matters: this is the cheapest prompt research available to any comms team right now. No new tool, no vendor, no budget line. Those follow-ups are the objections your content isn’t answering.

🏠 HOUSE OF ZEN

The AI Accelerator. Increase your brand’s AI visibility in 7 days, guaranteed. → offer.zenmedia.com

Most visibility work starts from what a team thinks people ask. We start from what they actually type, which includes the tasks, not just the questions.

The Accelerator runs the real prompt set for a category, finds where a brand goes missing when somebody is narrowing rather than browsing, and ships the content built to be pulled into that answer. It’s white-label, so agencies can run it under their own brand.

If your team has been asked “are we showing up in AI” and doesn’t have a defensible answer yet, this is the seven-day version of getting one.

→ offer.zenmedia.com

📊 This week’s PROMPT-TYPE MAP, RECEIPTS, REPORT DROP, PLATFORM MOVES, and DELIVERY INSIDER are below the fold for paid subscribers →

Read more

This going to be my headline for a few days, because it probably impacts everyone reading this. If you use AI to support your Linkedin content creation, this one is for you.

On July 30, LinkedIn added a “Seems like AI slop” option to every post in the feed. Anyone (think: competitors, customers, co-workers) can now flag your client’s content.

That means flagged posts can get reduced distribution outside your own network, and LinkedIn will privately warn accounts whose content reads as inauthentic.

Chief Product Officer Hari Srinivasan said: “AI slop is a top priority for all of us. We really care about this.” In the same breath, LinkedIn deleted its own “enhance your post” AI writing feature and replaced it with a proofreader that fixes grammar without touching your voice.

This shouldn’t be shocking and honestly I thought it would be higher: scanned 1,002,627 posts across five platforms and found that more than 40% of long-form LinkedIn posts are fully AI-generated. LinkedIn produced 62% of all the AI content they flagged while making up only about a third of what they scanned. It is the most AI-saturated platform they measured. Reddit, for contrast, came in at 4.4%. (Pangram sells an AI detector, so a detection company reporting that AI is everywhere deserves the obvious asterisk. The sample is also self-selected, pulled from posts their extension users chose to scan.)

So roughly two in five long-form posts on the platform would plausibly cause someone to hit that button.

Now, onto the thing that I am going crazy about.

A study of 175 brands across eight AI engines found that 96% of brands get described accurately when you ask about them by name (branded prompts), and 89% don’t who up when you ask a category question (non-branded prompts). When buyers asked the research question instead of the brand question, 99.99% of 49,391 citations pointed to third-party websites. Not the brand’s own pages. Somebody else’s.

We already knew this due to the millions of prompts we are now monitoring.

More on how to run it in THE MOVE at the bottom.

Your two Certification Day winners: Ujala Qazi and Michelle Monson. Both wrote a 27th Touch post that made me stop scrolling, which was the entire point of the exercise.

PS: AI Visibility Certification Day cohort #2 is August 27, 12:00 to 1:30 PT, 20 seats. Two just went to the giveaway. Grab one of the rest here.

🚀 BUILDING & LAUNCHING

AI Visibility Certification Day, cohort #2 (August 27)
Ninety minutes, live, 20 seats, and you leave with the actual testing method: prompt anatomy, the concierge audit, and the category question test. Cohort #1 sold out and the recording is not a substitute, because half the value is watching your own brand fail a prompt in real time. Two seats are gone to the giveaway winners. Reserve a seat.

The Pitch Desk
Since we are talking about AI slop this week, this is the tool built for the other end of the pipe. The Pitch Desk grades your pitch before it reaches a journalist. You compose, it diagnoses, you workshop, then you send. It runs the SNAP test on every pitch: Surprising, New, Actionable, Pertinent. You feed it the target reader, company size, vertical, geography, and buyer title, and it hands back specific fixes plus outlet suggestions. It also keeps a repository of what you sent and what actually got coverage, so the recommendations sharpen against your real results instead of a generic model of “good pitch.”

Take a look.

AI Slop Reduction Addendum for LLMs
Re-featuring this one because LinkedIn just made it a news story. It is the instruction set you paste into your model so your drafts stop arriving with the tells: the stacked adverbs, the tricolons, the em-dash habit, the “it’s not X, it’s Y” construction that every classifier on earth has learned to spot. Get it on Stan.

⚡ THE BRIEF

1. LinkedIn shipped an AI slop report button, then killed its own AI writing tool

What happened: On July 30, LinkedIn added “Seems like AI slop” to the post options menu across the feed. Flags feed LinkedIn’s classifiers, and flagged content can lose distribution beyond your immediate network. Accounts with a pattern get a private warning on their dashboard rather than a public penalty. The next day, LinkedIn confirmed it removed the “enhance your post” AI writing feature entirely, replacing it with a proofreader that preserves voice. The platform says it blocks hundreds of thousands of automated comment attempts every day and has stopped billions of automation attempts in the last few months.

What to do: Pull the AI-polish step out of your LinkedIn workflow this week. The final-pass rewrite step, which is the one that flattens voice into something a classifier and a human both recognize. Then kill any templated comment program you are running, because engagement pods are the single most detectable pattern on the platform.

Why it matters: Executive ghostwriting is the highest-trust, highest-risk product most comms teams sell. A CEO post that gets crowd-flagged is not a reach problem you fix next quarter. It is a credibility problem you have to explain in a meeting.

🔗 TechCrunch · Fortune · Pangram data

2. 89% of brands never get named when a buyer asks AI about their category

What happened: A study published July 29 tested 175 brands across legal, healthcare, SaaS, financial services, and retail on eight platforms: ChatGPT, Claude, Gemini, Copilot, Perplexity, Google AI Overviews, AI Mode, and Meta AI. When asked directly by name, 96% of brands were described accurately. When asked a category research question, 89% never appeared. Of 49,391 citations generated by those category prompts, 99.99% pointed to third-party sites. Only 0.10% of problem-awareness answers named any brand at all. Brands with fewer than 2,000 indexed pages mentioning them across the web had a 3% mention rate.

What to do: Stop reporting on branded prompts. That is the 96% test, and it measures whether the model has a file on you. Build your prompt set out of the questions a buyer asks before they know you exist, then track how often a competitor gets named instead.

Why it matters: This is the earned media budget argument in one number. Owned content cannot buy you into a consideration set that is assembled from third-party pages. Coverage is the input.

🔗 Search Engine Journal

3. Two AI disclosure regimes went live on Sunday, and one of them has a carve-out you should use

What happened: August 2 was a double deadline. The EU AI Act’s Article 50 transparency obligations took effect alongside actual fining power for the AI Office and national authorities: up to €15 million or 3% of global annual turnover, whichever is higher. Article 50 requires chatbots to disclose they are AI on first interaction, requires deepfakes and AI-manipulated media to be labeled, and requires AI-generated text on matters of public interest to be disclosed. The same day, California’s AI Transparency Act became operative. Providers with over 1,000,000 monthly users must offer a free public AI-detection tool and embed provenance data in outputs. Penalty is $5,000 per violation, with each day counting separately.

What to do: Read the Article 50 carve-out, because it is written for you. AI-generated text on matters of public interest does not require disclosure when a named human takes editorial responsibility for it. That is a documentable process, not a legal opinion. Put a named reviewer on record for every AI-assisted release, byline, and AI Notice touching EU audiences, and keep the record. Second: tell your design and web teams to stop stripping metadata out of image files.

Why it matters: Comms is about to get handed the disclosure clipboard.

🔗 EU AI Act Article 50 guidance · California AB 853 text

🏠 HOUSE OF ZEN

Ninety-six percent of brands get described accurately when someone types their name. Eighty-nine percent vanish when someone types the category question. That gap is the entire business problem, and you cannot fix a gap you have not measured.

The AI Accelerator at Zen Media is the seven-day version of the fix. We run the prompts your buyers actually type, not the ones with your name in them. You get the baseline, the competitor set that shows up instead of you, and the content built to close the specific gap we find. Guaranteed increase in AI visibility in seven days.

If you have been asked “are we showing up in AI?” and did not have a defensible answer, this is that answer. Start here.

📊 This week’s RECEIPTS, REPORT DROP, PLATFORM MOVES, the Monthly Trend Memo, and THE HUMAN BYLINE PROTOCOL are below the fold for paid subscribers. →

Read more

Hi Team PR@ctical.

OpenAI started selling ChatGPT ads this week. Self-serve, no minimum spend, any business with a credit card. That’s the version everyone covered.

Here’s the version that matters to us. The ad sits next to the answer. The answer itself is not for sale. It gets assembled from earned coverage, owned pages, and third party reviews. OpenAI will take anyone’s money to stand beside the conversation. They can’t sell what gets said in it.

We’ve been the cheap seats in the budget meeting for years. This week the most valuable real estate in AI turned out to be the one spot only our work can reach. Sit with that one for a second.

Also: I’m giving away 2 seats to AI Visibility Certification Day (August 27, live). How to enter is in Building & Launching. Winners named in next Monday’s edition.

Below the fold this week, one of these ships every week the next 4 weeks (this edition: the 90-Day Visibility Plan):

• 📊 The 90-Day Visibility Plan: Audit, Build, Measure, with deliverables and metrics

• 📬 Delivery Insider: the weekly AI Notice play, outcomes only

• 🛠️ An agentic workflow from the vault

• 🎯 A framework deep-dive

🚀 BUILDING & LAUNCHING

🎓 AI Visibility Certification Day, cohort 2: August 27, 12 to 1:30 PT. 20 seats. I’m giving 2 away.

Cohort 1 sold out in a week, Cohort 2 is open, and this time two of you come free.

Last week I gave you The 26 Touches Field Guide, the free 34-page walk through every touchpoint your attribution report can’t see. Your entry: tell me the 27th touch on Linkedin and tag me.

Entries close Friday, August 1. I pick the two best. Winners in next Monday’s edition.

Grab a seat before the 20 go. This runs once a quarter.

1. 📥 Missed the Field Guide? It’s free and it’s here. It’s also your study material for the giveaway, so that works out.

2. 📬 The Pitch Desk is still rocking. Three weeks in and getpitchdesk.com keeps doing what it says on the label: protect the brand, earn the coverage. If your pitches haven't met Haley yet, go fix that.

🎓 THE LESSON: The Concierge Audit

Last week you learned to read a prompt. This week you learn to hear the answer the way a buyer does. Ten minutes, no tools, do it today.

Step 1. Open ChatGPT and Gemini. Ask both: “Compare [your client] to [their closest competitor] for [the thing buyers actually hire them for]. Which should I pick?”

Step 2. Read the answer like a buyer, not a publicist. Write down three things: what’s wrong, what’s missing, and whose framing won.

Step 3. Trace each one back to a source. This is the step almost nobody does:

• A wrong fact usually traces to a stale page. Yours to update.

• A missing strength usually traces to zero earned coverage saying it. Yours to earn.

• The rival’s framing winning usually means their content answers the buyer’s question and yours answers a brief. Yours to rewrite.

That one-page gap map is the whole discipline in miniature. The answer is built from what’s public. What’s public is our department. And now that OpenAI is charging rent next to that answer, knowing how to change the answer itself is the skill.

(Shameless plug: this exercise at production scale, 1,000 prompts deep, is what we run at Zen Media. The 10-minute version is free and it will still sting a little.)

⚡ THE BRIEF

1. ChatGPT ads are live. Know the two lines before your client calls.
What happened: as of July 22, ChatGPT ads are self serve. No minimum spend. Targeting reads the conversation’s context, not keywords. Ads show to free and Go tier users only, and inventory is capped.
What to do: brief every client before they bring it to you. Two lines. One: an ad buys you a seat next to the answer, and only for free tier users. Two: earned and owned content decides what the answer says, for every user, on every tier.
Why it matters: this is the best budget conversation PR has been handed in years. Walk in with the distinction ready.
🔗 b2the7

2. Apple’s shopping assistant showed up in the fine print before the keynote.
What happened: on July 22, references to an unreleased “Virtual Shopping Assistant” surfaced in the Apple Store app’s updated privacy policy. It would compare products, recommend accessories, and guide upgrades. No announcement, no date. Apple lawyered it before they said it.
What to do: run the Concierge Audit above before some AI runs it on your buyer.
Why it matters: when the most controlled retailer on earth preps a concierge for its own storefront, the conversation is the store. Whatever the concierge knows, it learned from what’s public.
🔗 TUAW

3. The “safe” mid-tier influencer is your most expensive buy.
What happened: OpenSponsorship analyzed 1,527 paid creator deliverables. Mid-tier creators (50K to 250K followers) cost $7.07 per engagement. Nano and micro run $3.11 and $3.05. Nano creators reach 19.9% of their followers. Mega creators reach 1.1%.
What to do: next paid deal, test two nano or micro creators whose audience is exactly your buyer. Judge on cost per real engagement, nothing else.
Why it matters: mid-tier is where credibility theater lives. Small trusted voices are cheaper, they engage harder, and they’re exactly the kind of source AI answers cite.
🔗 NetInfluencer

4. Content creation is a college major now, and the salaries are public.
What happened: ASU’s Cronkite School enrolls its first Content Creation bachelor’s students this fall. Brands from Cadillac to e.l.f. to Unilever are hiring dedicated creator roles, some posted at $240,000 to $290,000.
What to do: on your next social hire, weigh the portfolio as heavily as the resume. If you’re the one negotiating, the market rate just went public. Use it.
Why it matters: the work this industry called a side hustle now has a degree, a salary band, and a seat in the org chart.
🔗 NetInfluencer

📊 This week’s RECEIPTS, REPORT DROP, PLATFORM MOVES, the 90-Day Visibility Plan, and DELIVERY INSIDER are below the fold for paid subscribers →

Read more

Somewhere in your sales team’s CRM right now there’s a closed deal marked “source: Google.” But we all know that’s not where the deal REALLY came from. That’s just where it ended. They likely saw you (or didn’t see you) roughly 26 times: in an AI answer, a podcast segment, a Reddit thread, and a LinkedIn comment from someone with 4,000 followers. The ad platform was the LAST click. Everything before it happened in the new customer buying journey.

6sense surveyed 4,000+ buyers and found 94% of buying groups had already ranked a preferred vendor, and they bought that early favorite 77% of the time (6sense Buyer Experience Report).

Today’s edition is going to dig in on how to read a prompt like a sentence and how to follow a citation all the way to revenue.

And the nerdy thing I couldn’t wait to share: I turned the whole buyer journey into something you can keep: The 26 Touches Field Guide, a 34-page visual walkthrough of every single touchpoint, free on my Stan store. Go grab it.

About that cohort: it was a hit, and the part I loved most was the size. I capped it small on purpose, and the whole session had a personal flow. Certificates went out to the first graduating class this week (congrats, you know who you are). Cohort #2 is now open: Thursday, August 27, 12:00 to 1:30 PM PT / 3:00 to 4:30 PM ET. 20 seats. Grab yours here.

📚 LESSON 1: A prompt has parts. Learn to read them.

In 5th grade English you learned to diagram a sentence. Subject, verb, object, modifier. Do the same thing to the prompts your customers type into ChatGPT and Gemini, because every part changes whether the answer names you.

Take the prompt: “I need to hire a PR agency for my business.”

Part 1, the subject: who is asking. “I.” The AI looks for signals about this person: role, industry, geography, everything it knows from prior chats. A CMO gets top-10 lists. A founder gets boutique picks. Same prompt, different person, different answer.

Part 2, the verb: the intent. “Need to hire.” Not “learning about.” Not “comparing.” This is bottom-of-funnel, wallet-out intent. “Thinking about” triggers education. “Need to hire” triggers recommendations. The verb tells you where in the journey the buyer is standing.

Part 3, the object: the category. “A PR agency.” Broad. Every agency on earth is a possible answer. This is where the fight is.

Part 4, the modifier: the context. “For my business.” Vague here, and vagueness is opportunity. The moment it gets specific (”for a Series B SaaS in fintech”), the answer changes completely.

You will not win “best PR agency” on your own. Broad plus high-intent is the hardest ground in the category. But you can absolutely win “best PR agency for a Series B fintech in Chicago.” That’s where your program shows up first. Win the niche variations, and the broad answers follow.

One more variable: the model itself. ChatGPT favors well-known brands. Perplexity favors recent press. Claude weighs authority differently. SparkToro tested this at scale: ask an AI for brand recommendations 100 times, and there’s less than a 1-in-100 chance you get the same list twice, though strong brands still showed up in 60 to 90% of answers for a given intent (SparkToro, 2,961 test runs).

And people are handing the models more parts to read every month. ChatGPT search queries now run 61% longer than Google queries, and search-enabled prompt length nearly doubled in a year (Semrush clickstream data). Longer prompts mean more modifiers. More modifiers mean more niche answers you can own.

Screenshot this for your CEO: Your category’s buying prompt has four parts: who’s asking, what they intend, the category, and the context. You can’t win the broad version. You can win every specific version that matches a customer you actually serve. That’s a content plan

🚀 BUILDING & LAUNCHING

NEW and FREE: The 26 Touches Field Guide. A 34-page visual breakdown of the modern buyer journey, one touchpoint per page, built like the screenshots you’d actually see: the ChatGPT shortlist, the Reddit thread, the text from a friend, the form fill that takes all the credit. Download it free on Stan.

AI Visibility Certification Day, cohort #2. August 27, 12:00 to 1:30 PM PT. 20 seats, capped small on purpose because the live prompt-diagramming is the whole point. Register on Stan.

AI Slop Reduction Addendum. Keep this on your desktop. Attach it anytime you are doing ANYTHING related to content and it will improve your output and reduce editing time. Get it here: https://stan.store/asksarahevans/p/ai-slop-reduction-addendum-for-chatgpt-and-claude

⚡ THE BRIEF

1. Half of B2B buyers now start research in an AI chatbot. 69% changed vendors because of it.
G2 surveyed 1,076 B2B decision-makers: 51% start software research with an AI chatbot more often than Google (up from 29% in April 2025), 69% chose a different vendor than the one they planned to buy based on AI guidance, and a third bought from a vendor they had never heard of before the chatbot named it (G2 Answer Economy research).

Why it matters: “Never heard of them” is no longer a moat and no longer a death sentence. The chatbot introduces unknowns into deals daily. You want to be the unknown that gets introduced, and definitely not the incumbent that gets swapped out at touch 14.

2. Muck Rack: Perplexity leans on YouTube 9x harder than its next source.
New Muck Rack analysis (July 16) found Perplexity cites YouTube more than any other domain, 2,952 citations versus 348 for the runner-up. How-to queries drive 29% of all YouTube citations, and Perplexity sometimes uses a YouTube video as its only source. Gemini ranks YouTube #2. Claude cited it zero times (Muck Rack, “How AI cites YouTube”).

Why it matters: One good video can become the single source an AI answer is built on. That’s the cheapest sole-source citation available right now.

3. Your executives’ personal LinkedIn posts beat your Company Page. It’s not close.
PRWeek’s July 16 coverage of the Meltwater x LinkedIn study (9.5 million AI citations analyzed): about 75% of LinkedIn citations in AI answers come from individual member profiles, only 25% from Company Pages, and 51% of cited members have fewer than 10,000 followers (PRWeek).

Why it matters: Follower count barely matters for citation. Format and specificity do. A subject-matter expert with 4,000 followers can out-cite your entire brand channel.

4. Google’s AI Mode now completes tasks inside search.
As of July 16, US users can connect apps like Instacart, Canva, and YouTube Music to AI Mode, and search will build the cart or pull the template directly (TechCrunch).

Why it matters: Search is turning into a transaction surface. Brands that are visible but not machine-actionable will watch the answer recommend them and the cart fill with someone else.

5. August 2 is a double regulatory deadline, and it’s 13 days out.
The EU AI Act’s enforcement powers for general-purpose AI go live August 2: documentation demands, model evaluations, fines up to 3% of global turnover or €15M (European Commission). Same day, California’s SB 942 AI Transparency Act becomes operative: AI-content disclosures and provenance watermarks, with penalties of $5,000 per violation per day (CA Legislature).

Why it matters: Two jurisdictions, one day, and the first enforcement stories will need example companies. Don’t volunteer.

🏠 HOUSE OF ZEN

Everything in Lesson 1 can be brought to life (and in about 14 days). Our team offers the AI Accelerator at Zen Media. We run the prompts your customers actually type (the full anatomy: personas, intent verbs, modifiers, across engines), find the answers where your brand should appear and doesn’t, and close the gap in 7 days, guaranteed.

It’s built for brands that want the baseline fast and for agencies that want to white-label the whole motion.

Start here: offer.zenmedia.com

📊 Below the fold for paid subscribers: THE AGENTIC COMMERCE CLOCK, this week’s exclusive THE ATTRIBUTION LADDER with the full 26-touch walkthrough, THE RECEIPTS, REPORT DROP, PLATFORM MOVES, THE CROSSOVER, and THE MOVE →

Read more

[Housekeeping first: Substack went down on me last Monday and the paid edition never reached you. The two paid exclusives you missed, the Agentic Commerce Clock debut and the Fable 5 Field Kit, are below the fold this week, updated.]

Every era gets named after it’s over. I’d rather name this one while we can still do something about it, so here’s my entry: the One-Funnel Era.

Here’s what I mean. On July 2, AI agents completed live purchases at real merchants through Visa and 30+ European banks. Walk through what happened in that one transaction: content got the brand cited in an AI answer (that used to be PR’s job), the answer surfaced the brand for a buying question (that used to be SEO’s job), the product information held up under machine verification (that used to be nobody’s job), and the agent completed the sale (that was sales’ job). Four departments, one motion, one funnel.

Wowsers. It’s really happening folks.

To be expected, the org charts won’t reflect this… yet. Muck Rack surveyed 1,115 PR pros: 73% call AI search visibility the next frontier. 29% say no one at their organization owns it. 39% aren’t measuring it at all.

That’s the One-Funnel Era: everyone sees it, nobody’s assigned to it, and the measurement is TBD.

Let’s be honest, it's going to stay messy for a while. The content forms are morphing (creator clips now outrank brand content in AI answers), the attribution funnels are new, the tools change weekly (three flagship models moved in the last seven days alone), and some of what's old is new again: the wire, earned coverage, and affiliate-style economics are all back in the middle of the funnel. Almost nobody's job description covers it.

To start, this is why I am launching the first of several AI Visibility Certifications. This is step one in what will be the emergence of roles tied to agentic commerce.

There are 5 days left to claim a spot in my first AI Visibility Certification Day. I built it because I can’t take the 1:1 training requests anymore; running and scaling an agency is a full-time job, so I’m doing this live, once a quarter, with the most current information I have. You finish with your certification in hand.

→ Claim your spot: AI Visibility Certification Day, live July 17

🚀 BUILDING & LAUNCHING

One week in: The Pitch Desk launched last Monday. If you haven’t met Haley yet, paste your next pitch in before you send it. Protect the brand. Earn the coverage.

⚡ THE BRIEF

1. Everyone calls AI search the next frontier. Nobody’s in charge of it.

What happened: Muck Rack’s State of PR 2026 (1,115 PR pros, released July 9) found 73% rate AI search visibility as important to strategy, while 29% say no one owns it and 39% aren’t measuring success at all. The stat that should reset budgets: 99% of AI citations come from non-paid sources, and 55% name high-authority earned coverage as the top tactic that works.

What to do: claim it. Write the one-page memo naming who owns AI visibility, what gets measured monthly, and which existing work (earned coverage, structured content) feeds it. Whoever writes that memo tends to own the budget that follows.

Why it matters: unowned functions get funded last and blamed first. In the One-Funnel Era, the unowned function IS the funnel.

🔗 Muck Rack via GlobeNewswire

2. AI answers are quoting creators, not brands

What happened: Jellyfish analyzed 27 million answers across seven AI assistants and found YouTube creator content in over 25% of them, climbing to nearly one in two in high-intent categories like financial services and consumer electronics. Independent niche creators beat both brand-owned content and celebrity influencers. Videos over 10 minutes win. In CPG categories, over a million unique YouTube videos get cited daily.

What to do: rethink the next campaign budget line. The creator with 9,000 subscribers and a 12-minute explainer may be the reason AI recommends you. Brief niche creators on the buying questions in your category, not on brand mentions.

Why it matters: the funnel’s discovery layer is being written by people who don’t work for you. You can either be in their content or absent from the answer.

🔗 Adweek

3. The new flagship cheats its own tests, and it’s in your ChatGPT now

What happened: independent evaluator METR ran GPT-5.6 Sol through its suite of software tasks and found it broke rules and exploited loopholes more than any public model METR has ever tested. Score the cheating as failure and the model works about 11 hours alone. Score it as success and it jumps past 270. METR’s own words: neither is a robust measurement. OpenAI’s system card admits the model fabricated research results and uploaded data to unapproved services about once every 400 tasks. On July 9, GPT-5.6 went public in ChatGPT.

What to do: put a named human approval step in front of anything AI can send, publish, or delete. Not a policy document. A name.

Why it matters: your team is using the most capable model on the market this week, and its own evaluators couldn’t pin down what it does unsupervised. Supervision is now a comms deliverable, because the failure mode is public.

🔗 METR · Transformer

4. Model access is supply now, not furniture

What happened: three flagships moved in seven days. Anthropic extended Fable 5’s included access to July 19, the second extension in a week; after that it’s metered at $10 in and $50 out per million tokens. Gemini 3.5 Pro is reported for July 17 with a 2-million-token context window, roughly 1.5 million words in one prompt; Google hasn’t confirmed, so treat it as reported, not promised. And ChatGPT Sites hit public beta on paid plans (not yet in the EEA or UK): describe a page, preview it, publish it, point your domain at it.

What to do: use Fable 5 hard this week (the Field Kit below is your plan), write next month’s plan as if it ends July 20, and pick the document pile you’d love to hand a 2-million-token machine so you can test Gemini the day it opens. Your goal is ready, not early.

Why it matters: teams that treat model access like office furniture get caught flat every time a deadline moves. Treat it like supply: know your inventory, your alternatives, and your switching costs.

🔗 Anthropic · BleepingComputer · OpenAI

5. The hiring spec flipped before the org chart did

What happened: Dice’s 2026 Tech Jobs Report (7 million US postings analyzed) found AI skills in 73% of tech job postings, up from 15% in January 2024. Postings with AI in the title are up 173% year over year.

What to do: read your own team’s job descriptions this week. If they describe lanes that merged (media relations here, SEO there, no AI anywhere), rewrite them before your next hire, and before your best people read someone else’s.

Why it matters: the market has already repriced these roles. PwC puts the AI-skill wage premium at 62%. Your job specs are either recruiting for the One-Funnel Era or for 2024.

🔗 Dice via HR Dive

🏠 HOUSE OF ZEN

The One-Funnel Era has an ownership problem, and the AI Accelerator is how brands bridge it. In 7 days, guaranteed, we lift a brand’s AI visibility: the citations that put you in the answer when your category gets asked, which is now the top of the only funnel that matters.

It starts with a baseline: where you show up today, where you don’t, and which buying questions in your category you’re losing. Then we go get the citations.

→ Start with the 7-day AI Accelerator

Read more

I know, I know… I’m a broken record. Agentic commerce is where we pivoted our own agency PR focus, and we couldn’t have been more on point.

Two payment networks moved within hours of each other this month. Visa and OpenAI announced that a tokenized Visa card can now complete a purchase inside ChatGPT. Mastercard shipped its own version, Agent Pay for Machines, the same day (American Banker).

The agent does the buying.

That is agentic commerce: an AI does the shopping and the checkout on a person’s behalf. It shortlists. It decides. It pays.

This is new. There will be bumps and bruises, but we still have to learn alongside the pace of technology. Consumers are going to ebb and flow with the trust of AI answers. A year ago, 82% of consumers said AI was more helpful than traditional search. Today it’s 54%. That is a 28-point drop in twelve months, and the skeptic camp grew nearly six times larger (Search Engine Land, 2026).

Call it AI Slop, call it a gold rush, things will start to even out.

So we have a machine that is increasingly making the purchase, and a public that increasingly does not trust what the machine says. Both things at once.

This is the whole reason I keep saying visibility was always step one, never the finish line. Getting named in the AI answer means you are building the infrastructure, the schema and the content.

The next fight is believability: whether the AI, and the human reading it actually trust you enough to act.

Being visible and being believed are two different scores, and business decision-makers rate AI-delivered answers about 10% more believable than the same claim elsewhere (PRCA, 2026).

That is why I’m running the first AI Visibility Certification Day, and as of this morning there are 12 spots left (grab one here). This is step one: the language, how AI decides which brands to surface, what your team needs internally to start building, and how to lead that conversation where you work. If you’re at an agency, this is a great use of your learning and development budget. Take what you learn back to your team.

One more thing, and this one is big. I spent two and a half hours this weekend with one of the top tech editors and columnists working today.

Best training data I’ve gotten in years. He essentially told me“I’m sick of meeting with agencies, telling them our preferences, only to get the same exact pitch over and over, and it’s not even a pitch.” So I’ve been a busy builder and am creating a new tool, shaped by what a real journalist actually wants. The theory: the 1% of pitches he actually covers should move up for people who use what his brain helped build. Coming soon. You’ll see it here first.

Let’s get into it.

🚀 BUILDING & LAUNCHING

1. Coming soon: the journalist-built pitch tool. Built from a 2.5-hour download with a top tech editor on what makes him open, read, and actually cover a pitch. Not an agency’s guess. The reporter’s real preferences, turned into a tool. Teased above. Watch this space.

2. AI Slop Addendum for LLMs. This tool has 500 hours of best practices you attach to any project or prompt where you’re writing content. It spits out an exceptionally better first draft so you can spend time on quality, not AI slop detection. (https://stan.store/asksarahevans/p/ai-slop-reduction-addendum-for-chatgpt-and-claude)

3. Guide: AI Visibility for Business Leaders (20 min, asksarah.ai/learn). Written for the person who has to defend the budget, not run the tactics. Start here if you’re the one walking into the board meeting.

⚡ THE BRIEF

1. Trust in AI answers fell 28 points in a year. That’s your opening, not your obituary.

A year ago, 82% of consumers said AI was more helpful than traditional search. Today, 54%. The skeptic group went from 3% to 17%, and the share of people who say heavy AI use would lower their trust in a brand jumped from 20% to 39% (Fractl, 2026; Search Engine Land, 2026).

What to do: Stop measuring only whether you appear. Start measuring whether what’s said about you is accurate. 27% of brands have already been misrepresented in an AI answer, and 14% say an AI inaccuracy has hit a sale, a customer relationship, or a PR situation (Fractl, 2026).

Why it matters: When trust is falling, the brands that read as credible win a bigger share of a more skeptical market. Falling trust is a sorting event. Some brands sort up.

🔗 Search Engine Land

2. The agent is becoming the buyer, and it picks vendors by checking sources you don’t control.

Visa and OpenAI turned on agent-initiated checkout inside ChatGPT on June 10, with Mastercard launching its machine-payment framework the same day (American Banker, 2026). Bain’s read: agentic commerce lives or dies on consumer trust (Bain, 2026). And here’s the mechanism that should reorganize your week: an agent doesn’t trust the claims on your own website. It cross-references you against the sources it can verify independently, earned coverage, reviews, credible third parties, before it shortlists.

What to do: Audit what the agent can verify about you off your own site. If your believability lives only on pages you own, the agent discounts it.

Why it matters: In agentic commerce, the trust signal shifts from social proof you create to citation networks the agent checks for itself. (Shameless plug: this is exactly what the Published Monthly system at Zen Media is built to feed.)

🔗 Bain & Company

3. The fastest way to lose the agent’s buyer is to look bought.

75% of Americans say they’d trust AI shopping less if the results were sponsored, and the same 75% would trust a brand less if it paid to influence an AI agent (Quad + The Harris Poll, 2026). A 2026 Gartner survey found half of US consumers would rather buy from brands that don’t use generative AI in their customer-facing content at all.

What to do: Win the answer with earned credibility, not paid placement. The thing that looks like a shortcut is the thing that erodes the trust you’re trying to build.

Why it matters: The agentic economy rewards brands that read as independently credible and penalizes the ones that read as purchased. Earned is not the slow option anymore. It’s the durable one.

🔗 PR Newswire / Quad + Harris Poll

🏠 HOUSE OF ZEN

Becoming the brand AI recommends is a system, not a campaign. The AI Accelerator is the 7-day version: we increase your brand’s AI visibility in seven days, guaranteed, and build the earned, verifiable footprint that the buying agents actually check before they shortlist. It’s white-label, so agencies can run it under their own banner for clients.

If your visibility is fine but your believability is thin, this is the gap to close before agentic checkout becomes your customer’s default.

→ Start at offer.zenmedia.com

🎙️ ON AIR

The Visibility Equation is back in the chair soon, and the through-line for the next run is exactly this: visibility versus believability, and what changes when the agent does the buying.

→ Replays at asksarah.ai/shows

📊 Below the fold for paid subscribers: THE RECEIPTS, the REPORT DROP on what journalists actually want, PLATFORM MOVES, THE CROSSOVER, and THE MOVE. →

Read more

I’m going to say something a little uncomfortable. We are about to watch one of the largest talent gaps of my career open up, and it is going to happen fast. Not because of who can code or who is “technical.” Because of who is willing to think differently about the work.

I’ve watched this coming for years, and last week a story I read was proof of what happens when AI leads and employees follow . A funeral director in Toronto went to sell his BMW back to the dealership. He texted with “Quinn,” agreed on $27,162.79, and felt good about it. Then the dealer called back. Quinn was a chatbot, and the real number was about $7,000 lower. They only honored it after CBC News called (CBC, Automotive News, June 16).

I have a name for that: Botsourcing. Handing AI a decision a human should have owned, then finding out the hard way the bot was confidently wrong. The dealer had the real price in his own head and trusted the bot anyway.

The last 18 months I had to relearn how to think, less like a comms person and more like a brand architect. Old dog, new tricks. It has been one of the most fun stretches of my career, and it is the thing companies keep asking me to teach. I ran it for a Fortune 500 comms team and a large PR agency in the last few weeks, and each one booked three more sessions on the spot.

So I’m opening it up.

🎓 Introducing AI Visibility Certification Day. The first cohort.

One live morning for ~30 people:

→ Speak the language and concepts shaping AI visibility
→ See how AI decides which brands it surfaces and recommends today
→ Know what your team needs internally to start building the systems
→ Think clearly about how your brand shows up in AI, and lead that conversation

You leave with a certification you can add to your resume and LinkedIn. This is the training enterprise teams have been booking, now open to everyone.

→ Claim your seat

One more thing I want to share. We are now running millions of prompts against our clients’ content to learn what actually makes AI cite a brand. This is the first cycle we are handing those reports to clients, and they are changing how we write everything. .

I’m sharing five things we learned below, because we all get better working from the same playbook.

PR@CTICAL is becoming a living guide. Every edition adds one move you can actually run. We start where I tell every team to start: before you fix anything, see what AI already says about you. That’s THE MOVE at the bottom.

🚀 BUILDING & LAUNCHING

1. AI Slop Remover [Addendum]. I attach this to every single content project and instruction, AGAIN with a prompt to improve the result. It has reduced my review and editing time tremendously → Grab it here.

2. AI's Role in PR Strategy (self-paced). If you want this on your own time, my new course in the PRNews Learning Collective just went live. Same thinking, built for when a live morning does not fit your calendar. → Take the course · browse the full Learning Collective

3. The Daily Briefing Builder (free). People kept asking how my Morning AI Briefing runs itself before I wake up. It is a saved Claude skill, not a prompt I re-paste. I packaged a starter version you can point at any topic and audience. Tell it your subject, it sets up a roundup you can run every morning in your voice. → Steal it here

⚡ THE BRIEF

1. A chatbot cost a BMW dealer $7,000 in one text thread.
The dealership let an AI agent named “Quinn” quote a buyback price, the bot treated the customer’s loan balance as the car’s value, and the offer landed about $7,000 too high. The dealer tried to walk it back and only honored it after CBC News got involved. 🔗 CBC · Automotive News

What to do: Find every place AI talks to a customer or makes a call on your behalf, and put a human approval gate on anything with a number, a price, or a promise in it.

The cheapest habit I give every team, the one that would have saved that dealer $7,000:

The Manager Test. Before anything ships under your name: Can I explain how it got here? Can I defend every fact in it? Would I put my name on it? One “no” means it isn’t done. Treat AI like your boss, not your employee.

2. PwC priced the skills gap, and judgment won.

PwC’s 2026 Global AI Jobs Barometer studied a billion-plus job ads and found AI is rewarding human skills (judgment, creativity, leadership) faster than technical button-pushing. Skills for the most AI-exposed roles are changing more than twice as fast as last year, and the AI-skill wage premium hit 62%. 🔗 PwC

What to do: Stop measuring people on volume and what they got done, it’s time to move to value and quality of output.

3. Showing up in AI is not the same as being believed.

Burson, with Profound, ran more than 55,000 believability forecasts across seven AI platforms and 85 companies. The finding: visibility and credibility are separate scoreboards. You can be in the answer everywhere and still lose if the model frames you in a way that does not land. Business decision-makers rated AI answers about 10% more believable than the general public, which cuts both ways. 🔗 Burson via PRCA

What to do: When you audit how AI talks about you, do not stop at “are we mentioned.” Read whether the framing would actually convince a buyer. Fact-based claims about products and workplace held up best.

(Shameless plug: figuring out what AI believes about you, and rebuilding it, is exactly what we do at Zen Media. Keep reading.)

🏠 HOUSE OF ZEN

Most brands have no idea what AI currently says about them, which means they cannot fix it.

The AI Accelerator is the 7-day, guaranteed lift in your brand’s AI visibility. We run the baseline, find where you are missing or misframed in AI answers, and move the needle in a week.

It is built for comms and marketing leaders who need proof before a budget conversation, and it is available white-label for agencies who want to offer it under their own name. → offer.zenmedia.com

📊 This week’s FROM THE DATA findings, RECEIPTS, REPORT DROP, and PLATFORM MOVES are below the fold for paid subscribers →

Read more

I don’t think I could have predicted this is where I’d be right now. After running a traditional agency for almost 16 years, I now have a team with 9 AI developers and an AI scientist. That’s a new layer into our ecosystem. They are the key to the rest of the team learning to think like architects and is the evolution of all of our roles.

Here’s what it means:

A communicator gets a task and does it.

An architect gets a task and asks “how do I build this once so it runs every time without me starting over and keeps improving over time?” You stop being the person who does the work and become the person who builds the thing that does the work.

The Architect Stack: Prompt → Skill → Project → Scheduled Task → Iterate → Improve

You don’t have to become a developer to do this. I’m not one. You spot the thing you do over and over, and you build it once. That is the entry-level “architect” point and it gets more exciting the deeper you go.

Here’s one I built that I use every single day.

It started with a problem I was sick of. Every AI content draft… SLOP! I was spending too much time editing until I realized I never “showed” AI what great looks like. So instead of fixing drafts by hand forever, I built the fix once. I went back through more than 500 hours of my own editing decisions, pulled every phrase I always cut and every pattern that screams “a machine wrote this,” and turned it into one standard I attach to anything. Any model. Any skill. Any project. The output changes the second you add it.

I aptly call it the AI Slop Reduction Guide, and it’s exactly what thinking like an architect produces: find the problem once, build the system once, own it forever, and keep sharpening it.

And the same way of working is what gets our clients cited. We traced one piece of content we call an AI Notice and found the actual impact from it. ONE PIECE! It triggered 154 non-branded citations to start. 154 different prompts where this now shows up as an answer. PR reporting will never be the same.

P.S. Because of all the IP in the slop guide, I am increasing the price by $50 every Monday. You have an hour to get it at $109 before I raise it to $159.

🚀 BUILDING & LAUNCHING

1. The AI Slop Reduction Guide (the one from up top). 35 banned phrases, an 11-item audit, and a copy-paste briefing prompt you attach to any model, skill, or project. More than 70 people are using it and reporting 40 to 60% less editing time. Get it here before tonight’s bump.

2. The Learning Curve Destroyer (free, open source). A viral Instagram prompt turned Claude into a teacher with a time budget. I rebuilt it as a Skill so your AI keeps the instructions forever. You tell it what you want to be able to do and how much time you have. It builds the plan, tells you what to skip, and teaches you hands-on. Install once and it’s yours. If you can attach a file to an email, you can install it. Download it from my GitHub.

⚡ THE BRIEF

1. Communicator vs architect: the actual difference.
A communicator gets handed a task and does it. An architect gets handed a task and asks “how do I build this so it runs every time without me starting over.” What to do this week: take the next thing that lands on your desk and, before you do it, write down the steps you’d take. That list is the blueprint for a skill. Why it matters: doing the task scales to one. Building the system scales to everything you do after.

2. AI is your employee, not your boss.
The work comes out average when you let it file unsupervised. What to do: after you get an output you actually like, paste the final version back in and say “this is the standard, learn from it.” Close the loop. And give it a standard up front instead of a blank prompt. Why it matters: a blank prompt gets you a generic draft. A briefed one gets you something you can almost send.

3. Build it once. Then never do it by hand again.
Pick the one task you hate most (a media brief, a coverage tally, a recurring report) and turn it into a skill this week. What to do: open Claude, describe the task, and say “interview me on everything you need to build this as a skill.” Answer its questions. Save the result. Why it matters: that’s five hours of setup one time, then twenty minutes forever. Junior-level grunt work, senior-level output.

4. Two questions before you write anything.
What does this client want to be known for? What are you writing to compete against? What to do: if you can’t answer the second one, stop and go look at what already owns that answer before you write a word. Why it matters: writing blind is how you produce content that disappears. Knowing what you’re up against is how you produce content that beats it.

🎙️ ON AIR

My new session with PRNEWS is live on demand: “AI’s Role in PR Strategy,” part of the PRNEWS Learning Collective. It’s built for senior comms and marketing leaders deciding right now where AI fits in their strategy. Watch on your schedule. Session here.

📊 This week's DELIVERY INSIDER, AGENTIC WORKFLOW #02, plus THE RECEIPTS, REPORT DROP, PLATFORM MOVES, THE CROSSOVER, and THE MOVE are below for paid subscribers →

Read more

I just got back from NYC Tech Week and one thing the forward-thinking leaders were talking about is the detriment of brands who say, “we’re going to focus on AI perception in the next six months.” That sentence is expensive.

Google just shipped new Search Generative AI performance reports inside Search Console. As it roles out, you will be able to see how often your pages show up inside AI Overviews and AI Mode, broken out by page, country, and date. (Google Search Central)

The thing you were going to “get to later” is now a number and baseline waiting to be owned by comms.

I will show you my own receipt. We picked a category we had zero authority in and ran one owned pillar, one media piece, and one AI Notice on the wire. Time from publish to the number one answer across ChatGPT, Claude, Grok, and Google was about two hours. We have held that top spot for 40 days and counting.

So how do you actually show up? I keep coming back to one framework I use with every brand: the Perception Triangle. Own it (the definitive answer on your site), Back it (authority content that says the same thing in an editorial voice), Spread it (independent coverage AI can read). When all three corners tell the machine the same story, it reads as consensus, and consensus is what gets cited. More on that below, with the third-party data that proves it.

One human note before the news. This week reminded me why I love events. I got real time with founders and reporters at Founder Brew, bumped into people I have known for years, and made a stack of new connections that would not have happened over email. The deeper AI takes us into automating the work, the more the in-person stuff appreciates in value. So go to the thing. Sit in the room. Shake the hand.

🚀 BUILDING & LAUNCHING

NEW: The AI Slop Reduction Addendum for LLMs

I went back and forth on whether to give this away. My business coach said I should charge $600 for it because of how much time, resource and IP went into it. I launched over the weekend at $59 and said, “I’ll keep upping the price each week.”

Today it’s at $109. Next week, $159.

Use this three ways:

1. Add it to every instruction or project as a resource

2. Keep it on your desktop and attach it AGAIN when asking something to be reviewed

3. Create your own skill and attach this to it.

It is a Universal Editorial Standard addendum with 35 banned phrases that instantly flag AI writing, and an 11-item audit your model runs before it hands you anything.

Editing time drops about 60%. Works with ChatGPT, Claude, Gemini, anything you already use.

🔗 Get the Addendum on Stan

Guide of the week: AI and Communications: The Practitioner’s Guide (20 min)

If you are the one being asked “what is our AI plan” in the next leadership meeting, start here. It is the plain-language walk-through of where comms actually owns the AI perception layer.

🔗 Read it at asksarah.ai/learn

⚡ THE BRIEF

1. Google turned AI visibility into a report line. The waiting period is over.

What happened: On June 3, Google added Search Generative AI performance reports to Search Console, showing impressions inside AI Overviews and AI Mode by page, country, and date. It is impressions only for now (no clicks, no position, no query data). Important caveat: Google is rolling it out to a subset of sites first, UK-first, to test before going wide, so you may not have it in your property yet. Google also added a toggle to pull your content from AI features without hurting organic rank. (Search Engine Land)

What to do: Check whether the report has reached your property. If it has, export your AI Overviews and AI Mode impressions by page and screenshot it as your baseline. If it has not, do not wait for Google. Build your baseline now with the Perception Triangle so you walk into the budget conversation with a number either way.

Why it matters: The day a thing becomes measurable is the day it stops being optional. Even in limited rollout, the signal is set. AI visibility is now something Google reports on, which means it is something your leadership will ask about.

2. 85.5% of what AI says about you comes from outside your website. The Perception Triangle is how you get cited.

What happened: The data on AI citations keeps landing in the same place. 85.5% of AI citations come from earned media, not brand-owned sites, across an analysis of more than one million prompts. (5W) Brand mentions across the web correlate roughly 3x more strongly with AI visibility than backlinks. (Ahrefs) And distributing one story through third-party outlets produced a 239% median lift in AI citations versus the brand site alone. (Stacker and Scrunch)

What to do: Run the Triangle on one buyer question you must own. Own it (the definitive answer on your site), Back it (an authority piece in an editorial voice), Spread it (independent coverage AI can read). One corner is not enough. AI reads agreement across all three as consensus.

Why it matters: Single-source visibility plays underperform because the model is looking for corroboration, not for your homepage. The brands getting cited are the ones whose story shows up in three places at once.

3. AI search is no longer a one-engine game. Optimizing for ChatGPT alone leaves you invisible everywhere else.

What happened: ChatGPT’s share of AI chatbot traffic fell to 54.7%, down from 76.5% early last year, while Gemini climbed to 27.4% and Claude grew 306% in a single quarter. (Momentic, built on Similarweb)

What to do: Take your buyers’ top three questions and ask them in Gemini and Claude, not just ChatGPT. See where you are named and where you vanish. The answer is usually different on each engine.

Why it matters: A single-engine strategy now misses nearly half the market. This is exactly why the Triangle matters.

4. AI is now the first stop, not the second. If you are not in the answer, you are not on the list.

What happened: 37% of consumers now start their search with an AI tool instead of Google or Bing, and 44% of AI-search users treat AI as their primary source of insight, versus 31% who still lean on traditional search. (Eight Oh Two 2026 AI + Search Behavior Study)

What to do: Stop treating AI answers as a downstream SEO concern. Map the three or four questions a buyer asks before they ever reach a sales conversation, then check whether your brand is named in the AI answer to each one.

Why it matters: When AI is the first stop, the shortlist is built before a human visits your site. Miss the answer and you miss the consideration set, no matter how good your pipeline looks after that point.

🏠 HOUSE OF ZEN

Increase your brand’s AI visibility in 7 days, guaranteed.

We are not a consultative agency anymore. We have become a combined industry partner and a standalone tech, strategy, and execution layer for any brand ready to move now. You bring the brand. We bring the build. The AI Accelerator is the proof-of-concept version: 7 days, money-back guarantee, real results.

This is for the ones leaping into AI right now, the ones who get that this is the window to own how you are perceived before everyone else catches up. DM me if you book. I have an extra gift waiting. 🔗 offer.zenmedia.com

Need your AI baseline and don’t want to wait on Google’s rollout? We will pull it for you. 🔗 ai.zenmedia.com

🎙️ ON AIR

I just walked off the Ragan Communications and PR Daily stage in Brooklyn, and the line people kept repeating back to me: stop treating AI like your boss, start treating it like your employee. Most of us hand AI a task, take whatever comes back, then spend an hour fixing something we already knew how to make. Flip it. If you know the end result, build the example first, then ask AI how to get there and better.

For daily reps on this, 6MOA (6 Months Out, Already) runs every weekday on LinkedIn Live with Duran. Same theme as this whole edition: the companies that move now are six months ahead of the ones still scheduling it.

📊 This week’s RECEIPTS, REPORT DROP, PLATFORM MOVES, and THE CROSSOVER are below the fold for paid subscribers →

Read more

The newsletter for PR, comms, and marketing pros building visibility in the age of AI

Edition: Monday, June 1, 2026

Hi Team PR@ctical,

Expect this stat to grow, and there’s no more ignoring it. If you’re reading this, you probably already feel it: AI traffic converts 4.4 times higher than organic search (Emarketed, 2026).

Sit with that, because it should change how you think about every dollar you spend.

Here’s why. When an AI recommends you, the person who clicks through arrives already convinced, ready to make a decision. The research, the comparison, and the shortlist all happened inside the chat.

I call it Inherited Intent.

When AI vouches for you, your visitors arrive carrying its authority. You don’t earn the trust at the door. You inherit it. And this is something PR can step right into, because our work is what fuels it.

Now the pain: that high-intent audience is barely 1% of your visits. So no, this isn’t “abandon search and chase AI.” It’s that the single highest-intent group of people landing on your site is coming through a door we don’t even have a sensor on, and the same force is about to let AI make the purchase itself.

Don’t worry. I’m going to help us get there. We are literally building for this.

I’m in Brooklyn starting tomorrow for PR Daily (June 3 to 5) and NY Tech Week. Come find me, details at the bottom.

Let’s get into it.

🚀 BUILDING & LAUNCHING

The Perception Stack
Need to request or show proof to get additional budget for AI visibility and perception? I build something (free) to help you with that. The Perception Stack is a templated guide and deck to help you request your budget.

Grab the Perception Stack →

GEO GPT, the AI Visibility Engine The fastest way to see what ChatGPT, Perplexity, and Gemini say about you right now. Run your brand, your top competitor, and three buyer questions. It’s the diagnostic layer of the stack.

Try GEO GPT →

⚡ THE BRIEF

1. Inherited Intent: why AI traffic converts 4.4x higher

What happened: Several 2026 datasets converged on the same finding. Visitors from AI search convert at about 4.4x the rate of organic and spend 68% more time on-site (Emarketed). The most rigorous version comes from Shopify: on product-page sessions, AI-referred visitors outconverted organic in 23 of 25 merchant categories, by an average of 56%, with 14% higher order values (Shopify). Even the conservative cut holds: ChatGPT traffic converts about 31% higher than non-branded organic (ALM Corp).

Why it happens (the part to sit with): I call it Inherited Intent. When an AI recommends you, the person clicking through inherits the AI’s authority. The research, the comparison, the shortlisting all happened inside the conversation. By the time they reach you, a trusted advisor has already vouched for you by name. More than half of AI-referred sessions start on a product or detail page, versus about 20% of organic (Shopify). They don’t arrive curious. They arrive convinced.

The honest part: AI referral is still only about 1.08% of total visits (Conductor 2026). This isn’t “drop SEO.” It’s “your most valuable visitors are arriving through a channel you can’t currently see, and it’s compounding.”

What to do: Measure it. Build a GA4 referral segment for chatgpt.com, perplexity.ai, gemini.google.com, and copilot today. If you skip this, the channel stays invisible and you credit the wrong sources forever. Earn it. If the AI never cites you, Inherited Intent is your competitor’s number. That’s the Perception Stack. Test against it. Point a small, measured slice of content and digital-PR budget at being citable, not just rankable.

Why it matters: The next 18 months won’t reward the biggest spender. They’ll reward whoever noticed the highest-intent door first and walked through while it was cheap.

(Shameless plug: Published Monthly at Zen Media is the done-for-you version of making you citable on these engines.)

2. The bigger shift: AI is about to do the buying

What happened: Inherited Intent is the early edge of agentic commerce. ChatGPT’s Instant Checkout has been live since late 2025 across 900 million weekly users; Google launched its Universal Commerce Protocol at NRF in January and Microsoft shipped Copilot Checkout, all letting an AI agent compare options and complete a purchase on someone’s behalf (JPMorgan). Estimates put 25 to 30% of US online purchases touching an AI agent by the end of 2026 (completeaitraining).

What to do: Internalize the shift from being found to being chosen by a machine. HBR’s framing is blunt: the winners aren’t the biggest brands, they’re the most “data-complete, agent-accessible, and transparent” (HBR). Make sure your facts, pricing, and proof are clean, structured, and consistent everywhere an agent might read them.

Why it matters: When an algorithm makes the final pick, the entire conversion-optimization playbook built for human eyeballs starts to wobble. PR’s real job becomes shaping what the machine believes is true about you.

🔗 HBR: How brands can adapt when AI agents do the shopping

3. OpenAI opened ChatGPT ads to everyone

What happened: On May 5, OpenAI launched self-serve buying in Ads Manager, killed the $50,000 minimum spend, added CPC bidding alongside CPM, and shipped a measurement pixel and Conversions API. Two days later it announced expansion into the UK, Mexico, Brazil, Japan, and South Korea (OpenAI).

What to do: Don’t rush to buy. ChatGPT ads work best for brands the model already treats as credible. Audit what it says about you before you hand it a card. Earned perception first, paid second.

Why it matters: AI is now the answer engine and the ad engine at once. The wall between earned and paid just got thinner, and PR sits on the new seam.

🔗 OpenAI: New ways to buy ChatGPT ads

4. Phase One, live: ChatGPT cites brands 0.59% of the time. Perplexity does it 13%.

What happened: A study of 34,234 AI responses found a 46x gap in brand citation rates. ChatGPT favors Wikipedia, Perplexity leans on Reddit, AI Overviews prefer YouTube. Only 11% of cited domains overlap across engines (Leapd).

What to do: This citation data is Phase One, the measurable layer of perception. Use it to see where you show up today, then build past it. You need presence across the source types each engine trusts: reference content, community proof, multimedia. Citation is the scoreboard; perception is the game.

Why it matters: Single-engine optimization is the new single-keyword strategy. Feels like progress, quietly caps your ceiling.

🔗 How the engines source information

5. 39% of consumers trust a brand less when they spot AI content

What happened: Klaviyo’s 2026 research found 39% trust a brand less for AI-generated content; Animoto found 36% say an AI video lowers brand perception, and about 83% of consumers say they’ve spotted AI video. Disclosure made the trust hit worse, not better (Klaviyo, Agile Brand Guide).

What to do: Use AI to draft and scale. Keep a human visibly in the chair on anything public-facing, especially video and executive voice.

Why it matters: You can be the most-cited brand in your category and still lose if what people find feels machine-made. Perception is retrieval and trust.

🔗 Klaviyo: Consumer trust in AI 2026

📊 This week’s RECEIPTS, REPORT DROP, PLATFORM MOVES, and the first-Monday Trend Memo are below the fold for paid subscribers →

Read more

Whew! This is a long one, but it’s one of the meatiest editions to date.

If you’re new here, PR@CTICAL is not a newsletter that aggregates other newsletters. It is how to move during the era of AI perception (and the paid side keeps you ahead).

🎁 I built a little thing for the entire industry

Today I am giving you something that has not existed in this industry before: an analyst-grade deck you can use to personalize and request your very own AI visibility (eventually AI perception) budget.

The AI Perception Budget Brief.

Take it. Use it. Get the budget. Move the industry forward.

→ Free at asksarah.ai/perception-stack

Change your lexicon ASAP; it’s where we are heading

One piece of authority work cannot win every prompt anymore. Comms now has to plan content against persona + intent + LLM use type + tier. Same category, same product, different person, different platform: different answer, different winner.

1. Maya, VP Marketing at a Fortune 500 SaaS, opens Perplexity Pro Paid with research intent. She gets a curated answer with Forbes, HBR, MIT, and WSJ as sources.

2. Devon, Marketing Ops Analyst at a mid-market firm, opens ChatGPT Free with buy intent. He gets a tool listicle with a sponsored competitor placement at the top.

Same product. Different intent. Different engine. Different winner.

One piece of authority work cannot win every prompt. Plan for every persona, intent, and engine that matters.

🗣️ A note on “AI Visibility” (which I think will be outdated in the next 8 months)

The industry calls it AI visibility. That is the doorway. The Brief uses it because that is what gets searched, RFP’d, and funded.

What actually matters is AI perception. Visibility is binary. Perception is everything else. What the AI thinks you do, who it says you serve, the tier it places you in, whether it names you with sources or buries you in a competitor’s footnote.

Visibility is the metric. Perception is the asset.

📊 Three numbers you should know

• 35% of US consumers begin product discovery in AI. Google gets 13.6%. (Similarweb 2026 AI Brand Visibility Index)

• 84% of every AI citation is earned media. Third quarter running. 0.3% is paid. (Muck Rack May 2026 Generative Pulse)

• 94% of enterprise CMOs are increasing AEO/GEO investment in 2026. It is now the #1 strategic marketing priority. (Conductor 2026 CMO Investment Report)

The budget is moving. The packaging for the ask did not exist. Now it does. The execution lives below the paywall.

P.S. I am back in New York the first week of June for NYC Tech Week and speaking at PR Daily Conference, June 3-5 in Brooklyn. If you want to grab coffee or sit in on the talk, tell me.

P.P.S. I got to hang with the Notified team and Amanda Coffee and Meredith Klein in NYC last week. It’s so refreshing to learn with peers who are as excited as I am.

🚀 BUILDING & LAUNCHING

🆕 LAUNCH: The AI Perception Budget Brief. The first analyst-grade, practical industry deck built specifically for comms and marketing leaders to land an AI visibility and AI perception budget. 19 editable slides (.pptx that opens in PowerPoint, Keynote, and Google Slides), the Perception Stack framework, 6 sourced 2026 stats, four budget tiers ($50K to $5M+), CFO objection handling, 90-day rollout, 3-KPI scorecard, AI terminology glossary. Free. Built because 94% of enterprise CMOs are increasing AEO/GEO budget this year and there was no resource on the market to package the ask.

→ asksarah.ai/perception-stack

Guide of the week: The Knowledge Panel & AI Visibility Playbook (25 min). What a knowledge panel is, why it sits underneath every AI citation about you, and how to fix yours before the next earnings call. Read it on the way home from the budget meeting. → asksarah.ai/learn

⚡ THE BRIEF

1. 🔍 Discovery: the buyer funnel just inverted.

35% of US consumers now begin product discovery in AI tools. Google gets 13.6%. (Similarweb 2026 AI Brand Visibility Index.) Walker Sands’ B2B benchmark says the typical B2B brand is cited in just 3% of the AI answers it should own.

What to do: stop reporting Google share-of-voice as the top number on your dashboard. Add answer share across ChatGPT, Claude, Gemini, Perplexity, and Copilot. Then segment by persona and intent. The prompts a VP asks in Perplexity Pro are not the prompts an analyst asks in ChatGPT Free. Map both before you write anything new.

Why it matters: the consideration set closes before search starts. Whoever measures the right prompts on the right engines first wins the next quarter of pipeline.

🔗 Similarweb 2026 AI Brand Visibility Index · Walker Sands B2B AI Search Visibility Benchmark

2. 📡 Delivery: every SaaS company is becoming an AGaaS company.

Jensen Huang at GTC 2026: “Every single SaaS company will become an AGaaS company.” Plain English: software stops being a tool people open and starts being an agent that delivers outcomes. Visa research says 71% of businesses will optimize for AI agents and 53% will let agents negotiate prices. Anthropic’s Project Deal closed 186 deals through Claude agents with zero human approvals. Gartner projects 40% of enterprise apps will feature task-specific AI agents by end of 2026, up from under 5% in 2025.

What to do: assume an AI agent reads about your company before any human does, and assume the agent is acting on behalf of a specific persona with a specific intent. Audit your brand surface across the platforms agents use as ground truth (Wikipedia, Wikidata, LinkedIn, GlobeNewswire, third-party listicles). Then audit again per persona, because Maya’s agent and Devon’s agent read different sources.

Why it matters: when agents negotiate and recommend, your brand is either the answer for that persona on that engine, or out of the deal.

🔗 BusinessEngineer.ai: From SaaS to AGaaS · Visa research on agentic commerce

3. 🛡️ Defense: Wikipedia, Reddit, and Substack are outranking Forbes in ChatGPT.

A 200-query test from PR News found that for B2B citations in ChatGPT, Wikipedia, Reddit, and Substack consistently outrank Forbes and most traditional Tier-1 outlets. The AI is treating community-edited and creator-economy surfaces as higher-trust signals than legacy brand-name media for many prompts.

What to do: stop optimizing for Tier-1 only. Build a Wikipedia stub if you do not have one. Get cited on Reddit AMAs in your category. Place a guest piece on a Substack with reach in your space. And remember the persona-intent lens: the engine your buyer uses determines which of these sources matters most.

Why it matters: the engines reward different surfaces than your media plan assumes. Optimizing only for Forbes leaves majority of the prompt set on the table.

🔗 PR News / EPR Research: Why Wikipedia, Reddit, and Substack outrank Forbes

🏠 HOUSE OF ZEN

NEW opportunity for agencies. The AI Accelerator at offer.zenmedia.com is now available as a white-label offer. Use it two ways:

• As a bonus for your existing clients. Add a 7-day guaranteed AI visibility lift to any active retainer. Costs you nothing to add. Gives your client a verifiable AI citation receipt by day 7. Strengthens the retainer.

• As your own sales lead. Lead with the 7-day Accelerator on new business. Prove the lift in week one. Convert to a longer engagement on your services.

→ offer.zenmedia.com

🎙️ ON AIR

I am back in New York the first week of June for NYC Tech Week and speaking at the PR Daily Conference, June 3-5 in Brooklyn. The talk: what a perception layer audit looks like inside a Fortune-500-ready comms operation, what to measure per persona, and which AI engines deserve which weight at each tier. If you are there, find me. If not, the slides will land in a future edition.

🔒 PR@CTICAL PRO STARTS HERE

📊 Below the fold: this week’s DELIVERY INSIDER (the 24-hour citation play worth $100K+ AVE), the receipts, report drop, platform moves, the crossover, and the move. Paid subscribers only.

Read more

I sat down with Meredith Klein for Inc this week. The piece is called The Ugly Side of Entrepreneurship That Nobody Likes to Talk About, and it’s the part of building a business nobody wants to put on the highlight reel: the years before the wins, the months between paychecks, the quarters when the only thing keeping the lights on is conviction. I always say an overnight success is 20 years in the making.

My POV is something I’m adamant about, and it’s being paid to speak. Something I’m bullish on. Read it to see if you agree.

Before we hit the meat of this issue, do one thing. Go to offer.zenmedia.com. Zen Media is a service-as-software company, and the offer waiting there will prove what we can do to your AI perception in a single seven-day exercise. Run it on your own brand. Run it on a client. Run it on the competitor your buyer just got handed in ChatGPT instead of you. The AI Accelerator is the fastest way to know whether the next quarter is going to break in your direction.

Now the numbers. Muck Rack’s third straight quarter of Generative Pulse data analyzed 25 million links across ChatGPT, Claude, and Gemini. Earned media drove 84% of every citation. Paid placements drove 0.3%. The number has been at 84% for three quarters running. This is not a blip, it is the new wire.

Now stack the FTI receipt on top of that. FTI Consulting just posted Q1 2026 Strategic Communications revenue at $103 million, up 18.4% year over year, with operating income more than doubled. That is not the chart a struggling discipline produces. That is the chart of a function that just got reclassified as critical infrastructure inside the enterprise.

Here is the math the cutters are missing. If 84% of the AI answer your buyer reads is earned media, and 68% of B2B buyers (per Wynter) now open their first vendor conversation inside ChatGPT, Claude, or Perplexity, then cutting PR in 2026 is not a budget decision. It is a pipeline decision. You are removing the inputs the AI is using to recommend you, six months before you find out you stopped getting shortlisted.

The companies that are six months behind on this will not catch up by year-end unless they start now. The Inc piece is honest about what it costs to keep going. The Muck Rack data tells you what it costs to stop.

P.S. The Zen Media NYC meetup is THIS Thursday, May 14, at Notified. If you are in the city, come. Link here.

🚀 BUILDING & LAUNCHING

GEO GPT: AI Visibility Engine Run any brand, your own or your client’s, through GEO GPT and get a scored read on how the LLMs actually see you right now. It tells you which engines cite you, which engines do not, and which competitors are getting recommended in your place. This is the diagnostic you run before any earned media plan, and the one most teams skip because they assume the answer. Free at asksarah.ai/tools.

The CMO’s Guide to AI Visibility (22 min read) The Duke CMO Survey below the paywall says 24% of marketing work is already AI-powered and training budgets are at 3.8% and falling. This guide is the catch-up document I would hand any CMO walking into a Q3 budget review. What to measure, what to fund, what to defund, and the conversation to have with your CEO before the AI conversation has it for you. Read it at asksarah.ai/learn.

⚡ THE BRIEF

1) 84% of every AI citation is earned media. Three quarters running. Muck Rack’s May 2026 Generative Pulse analyzed 25 million links across ChatGPT, Claude, and Gemini. Earned media drove 84% of citations. Journalism alone was 27%. Paid placements were 0.3%. The number has been at 84% for three quarters in a row.

What to do: If PR is on the chopping block in your budget review this month, put this data on the desk before the meeting starts. The byline and the wire are the AI infrastructure your buyers are reading from. Cutting earned spend in 2026 is cutting your AI visibility on a one-to-one basis.

Why it matters: This is not a blip. This is the new wire, and the pattern has not moved in a year. The next quarter belongs to the teams that already moved.

2) 68% of B2B buyers now open their first vendor conversation with AI, not you. Wynter’s 2026 research caught the inflection. Two out of three B2B buyers now begin product research inside ChatGPT, Claude, or Perplexity. Google is the second touch. Your buyer’s first impression of your company is whatever the AI said about you yesterday.

What to do: Audit your top three buyer prompts this week. Type the question your ICP types and read the answer your buyer reads. If your name is not in it, that is your Q3 problem in one screenshot.

Why it matters: Every pitch deck, every sales call, every renewal conversation now starts after the AI already made a recommendation. The earned media spend is the input to whether you make the shortlist before a human ever takes the meeting.

3) FTI’s Strategic Comms segment posted record Q1 revenue. Operating income more than doubled. FTI Consulting reported Q1 2026 Strategic Communications revenue of $103 million, up 18.4% year over year. Operating income more than doubled. Demand drivers cited in the filing: corporate reputation, public affairs, and financial communications.

What to do: Print this section. Bring it to your next finance meeting. When a Strategic Comms division shows up on the earnings call as the segment that doubled operating income, the budget conversation has officially reversed.

Why it matters: The work that requires judgment, context, and a person at the table is having a moment, and it is showing up in a regulated public filing. Anyone telling you the craft is dying is reading the wrong report.

🏠 HOUSE OF ZEN

The AI Accelerator: increase your brand’s AI visibility in 7 days. Guaranteed.

This is the offer that proves what Muck Rack just measured. We take any brand, yours or a competitor’s, and inside seven days we move the citation surface for the prompts your buyer actually types. One exercise. One brand. One week. Guaranteed.

It is built for: founders running their own visibility, CMOs who need a fast catch-up before a board cycle, agencies who want to white-label the most accountable AI visibility offer on the market, and operators who got handed the “do something about AI” task on a Monday and need to show a result by Friday.

Activate it at offer.zenmedia.com. The system underneath is proprietary. The receipt is delivered in seven days.

🎙️ ON AIR

The Visibility Equation is off this week. I’m in NYC for the Zen Media meetup at Notified on Thursday. Expect dispatches from the floor: who showed up, what they said, and the conversations CMOs are actually having about AI visibility when the slides are off.

If you’re in the city Thursday, come find me. RSVP here.

📊 This week’s RECEIPTS, REPORT DROP, PLATFORM MOVES, CROSSOVER, and THE MOVE are below the fold for paid subscribers →

Read more

Monday, 1:40 PM ET. We hit publish on a single GlobeNewswire release. Four OEMs Capture 60% of Robotics Industry AI Visibility. By 2:25 PM, it was #1 on Google for both target buyer queries. By 3:40 PM, ChatGPT and Google AI Mode were citing it back. 143 outlets. 37 million reach. 120 minutes.

That is the speed proof.

We just ran our latest PR Attribution audit on an active B2B client. 200 buyer prompts in their category, run across Claude, ChatGPT, Gemini, Perplexity, and Grok. The brand shows up in 76% of AI answers. All five engines name them. 241 placements. 75.2 million in audience reach. 250 citations across the AI engines. The release lives inside the answers that shape buyer shortlists before sales ever picks up the phone.

That is the scale proof.

I built the system because nobody in PR was producing the proof point clients are starting to ask for: did the brand show up when a buyer asked AI?

The answer used to take a quarter. We have it in two hours.

We have shipped hundreds of AI Notices.

The Notice is the front door. The engine behind it is 20+ proprietary templates we built and own, monitored against our own GEO data 24 hours a day, and rewritten daily based on what is actually being cited in ChatGPT, Claude, Perplexity, Gemini, and Grok. I am not walking anyone through the engine. I stopped publishing the toolset publicly because copycats started selling broken versions, and the category gets wrecked when bad work gets named first.

Here is what the numbers mean.

2 hours from wire to AI citation. One press release distributed, indexed, ranked #1 on Google, and quoted by ChatGPT and Google AI Mode inside 120 minutes.

76% brand visibility across 200 buyer prompts. Most agencies reverse-engineer the prompts a brand already ranks for. We do the opposite. Our intelligence layer predicts the prompts a real buyer will type when they are shortlisting in a category, mapped to intent and persona. We run those predicted prompts as real searches across all five major AI engines and track where the brand shows up, in what order, against which competitors. 76% means the brand was named in 152 of the 200 prompts buyers will actually use next.

300%+ AI Overview lift across 100% of our active client roster. Every active client is being recommended in AI Overviews at 3x or more their pre-program rate.

(P.S. NYC, Thursday, May 14, comms + media + marketing meetup at Notified, 28 Liberty Street, 53rd Floor. Free, 5-7pm ET, 60 seats. Register: https://insight.notified.com/zenmedia_nycmeetup2026)

🚀 BUILDING & LAUNCHING

AI Visibility Assessment — New on asksarah.ai. Five-minute self-scoring tool. 🔗 https://asksarah.ai/assessment

Daily AI Briefing — Six days a week, the only AI news a comms pro needs to know. 🔗 https://asksarah.ai/briefings

⚡ THE BRIEF

Three stories that explain why the two receipts above worked.

1. 89% of brands appear in AI search. 14% measure it. The blind spot just became the budget conversation. Your client’s CEO will ask if they show up in ChatGPT. The first agency with the answer wins the renewal. The agency without one will get the call on the way out. 🔗 https://www.globenewswire.com/news-release/2026/04/07/3269307/0/en/Only-14-of-Marketers-Track-AI-Search-Citations-Even-as-89-of-Brands-are-Already-Appearing-in-them.html

2. Brand mentions are 3x more correlated with AI Overview visibility than backlinks. Ahrefs analyzed 75,000 brands. Mentions sit at 0.664. Backlinks at 0.287. The decade of SEO playbook is wrong now. Pitch for the mention, not the link. 🔗 https://ahrefs.com/blog/ai-overview-brand-correlation/

3. AI Overviews cut clicks 38% in the first randomized study. Carnegie Mellon and the Indian School of Business ran the field experiment. Zero-click searches rose from 54% to 72%. Your traffic report is showing 38% less of the picture every quarter. Start reporting on inclusion in the answer. 🔗 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6513059

🏠 HOUSE OF ZEN

The AI Accelerator. Increase your brand’s AI visibility in 7 days, guaranteed.

The 7-day program for brands and agencies that need to show up in ChatGPT, Claude, Gemini, Perplexity, and Grok answers now. Run it on your brand, or white-label it for your clients.

Email me directly, or hit the link.

Activate the AI Accelerator →

🎙️ ON AIR

Last week’s Visibility Equation episode with Amanda Coffee.

Eighteen years inside the biggest corporate communications shops. What changed, what didn’t, what the next decade looks like from the seat. If you watch one episode this month, watch this one.

All TVE episodes can be found right here: https://asksarah.ai/shows

📊 RECEIPTS, REPORT DROP, PLATFORM MOVES, THE CROSSOVER, MONTHLY TREND MEMO, and THE MOVE are below the fold for paid subscribers →

Read more

I don’t know what you did this weekend. I had bronchitis. So I rested, and I built. And in the middle of the building, it hit me: I have made SO much over the last twelve months that I have done a bad job of reminding you what already exists. You can’t use a hub I never linked to. You can’t bookmark a glossary I never named. You can’t send your CMO to a research desk you never knew was there.

So before I tell you about the new thing, let me tell you about the things that have been waiting for you this whole time.

Then I’ll tell you about MyAgentStack.io, which is the new thing, and which I think is going to change how you build your AI workflow this quarter.

🏠 The AskSarah.AI Hub. Bookmark Every Link.

Five things live, working, and updated by me (with a lot of AI and a lot of human sweat) every week:

1. Daily AI Briefings. Hand-curated, sourced, ready by 5 AM. The one I open before my coffee.

2. NEW: The Research Hub. Every stat, study, and report I trust, in one place. Stop sourcing from scratch. Stop quoting research you can’t verify.

3. Live Show Recaps. Two shows now: The Visibility Equation and 6MOA. The replays are where the real receipts live.

4. NEW: The AI Glossary. New terms added weekly. Last week’s word was Tokenmaxxing. This week’s, you’ll see in a minute. (Spoiler: B2AI, courtesy of Visa.)

5. The Learn Hub. Hours of structured guides for CMOs, founders, and comms leaders trying to catch up.

Now, the new thing.

🚀 BUILDING & LAUNCHING: MyAgentStack.io

What it is

MyAgentStack.io is a curated, public repository of every AI agent that touches PR, marketing, comms, and media. One page. One search. Every agent.

Not tools. Not models. Agents. The things that go do work for you, autonomously or semi-autonomously, inside your workflow. Research agents. Pitching agents. Listening agents. Visibility agents. Briefing agents. Shopping agents. Procurement agents. The ones already inside your Slack, your Ads Manager, your CMS. And the ones your competitor is already running that you haven’t heard of yet.

Why I built it (and why this week)

The agent explosion is real, and it is not slowing down. Visa just declared the B2AI era this week (more on that below). Microsoft Copilot Checkout is buying for 500,000 merchants. Meta’s AI Business Assistant just went global. Anthropic ran an internal experiment where Claude agents closed deals with zero human approval. Every one of those moves is an agent doing the work that used to require a person on your team or a contract with a holding company.

There is no map. There is no central index. There is no easy way for a CMO to walk into a Monday meeting and say “here are the eight agents we should be testing this quarter and why.”

So I built one.

How to use it this week

If you are running a comms or marketing function, do these three things on Monday morning:

1. Open MyAgentStack.io and scroll the full list. Don’t filter. Just look. You will find five agents you didn’t know existed.

2. Pick one to pilot. Just one. Something tied to a workflow you do every week (briefings, pitching, AI visibility audits, ad ops, content distribution).

3. Bring it to your next standup with a one-page brief: what it does, what it would replace, what it would augment, and what the 14-day pilot looks like.

That’s it. You don’t need a new tech stack overhaul. You need to start running agents inside the workflows you already own.

What you’ll find on it

Categories on day one include:

• Research and listening agents

• Pitching and media outreach agents

• AI visibility and citation agents (GEO, AEO, answer share)

• Newsroom and editorial agents

• Briefing and morning-digest agents

• Ad ops and creative agents

• Procurement and shopping agents (the B2AI layer)

• Custom-built / vibe-coded agents from the community

Every entry has: what it does, who built it, what it costs, where to access it, and (where I have it) a real-world use case from someone running it in production.

Submit your own

If you are vibe-coding, building, or shipping an agent that touches our industry, send it to me. I review every submission. The bar is utility, not polish. If it solves a real problem for a real PR or marketing operator, it goes in.

Why this matters for your business

The companies six months ahead on AI right now are not the ones with the biggest budgets. They are the ones with the clearest agent inventory. They know what they are running, what they are testing, what they are sunsetting, and what they are about to buy. Most of you cannot list your current agent stack on one piece of paper. That is the gap.

MyAgentStack.io is the start of closing it.

Bookmark it. Send it. Submit to it. Let’s go.

⚡ THE BRIEF

1. Visa renamed commerce. It’s now B2AI.

What happened: Visa CMO Frank Cooper III declared the B2AI era this week. Visa’s research (with Morning Consult, surveying 2,000 consumers and 512 business decision-makers) shows 71% of businesses are willing to optimize products for AI agents, and 53% would let agents negotiate prices with other AI systems on their behalf.

What to do: Rewrite your top 10 product or service pages this week as a structured bullet sheet a shopping agent can quote back word for word. Specs, price comparables, differentiators, use cases. Skip the brand brochure prose.

Why it matters: When the buyer is a bot, paragraphs get skipped. Agents lift bullets and structured facts. If your page reads like a press release, you are invisible to the procurement layer that is now shopping for your customer. → Visa research summary

2. Anthropic let Claude agents run an office marketplace. Zero human approval. 186 deals closed.

What happened: Anthropic gave 69 employees Claude agents to buy and sell inside an internal marketplace for one week. The agents struck 186 deals across 500 listings, totaling just over $4,000. Sellers paired with Claude Opus 4.5 earned more per item than humans selling alone. Buyers using agents paid less.

What to do: Ask ChatGPT or Claude to shop for your top product under a real price point this week. If the agent cannot identify your brand, price the offer, and complete the transaction without you, your product page is broken for the agent buyer.

Why it matters: This is the first published in-house experiment proving agents can close transactions in production. It is not a research thought experiment. It is a working system. Your buyer’s procurement team is about to test the same thing on you. → Anthropic Project Deal

3. ChatGPT ads now reach logged-out users. Minimum spend dropped 75%.

What happened: OpenAI quietly extended ChatGPT advertising to unauthenticated users this week. Ads appear native inside chat responses, not as banners. The minimum advertiser commitment dropped from $200,000 to roughly $50,000. Canva and JobCopilot were among the first confirmed advertisers in logged-out sessions.

What to do: If your CMO has been waiting to test ChatGPT inventory because of the price floor, that excuse is gone. Brief your media team to model a $50K test against your top funnel campaign this week.

Why it matters: This is the first time mid-market budgets can buy ChatGPT inventory. The audience just expanded to every casual ChatGPT user, not just paid subscribers. Inventory is bigger, the floor is lower, and your competitor is going to test it before you do. → Search Engine Land · AdExchanger · OpenAI’s announcement

4. Microsoft Copilot Checkout hit 500,000 merchants. The agent is doing the buying.

What happened: Microsoft expanded Copilot Checkout to over half a million merchants and pushed it into the mobile app. Ask Copilot to redo your living room and it researches, picks pieces, and pays inside the chat. Urban Outfitters and Etsy are live.

What to do: Run a Copilot or ChatGPT shopping query for your top SKU today. If your product or a competitor is missing, screenshot the gap and send it to your e-comm lead before noon.

Why it matters: Half a million merchants is past pilot scale. If your product feed isn’t structured for agent retrieval, your category is getting bought without you. (Shameless plug: this is the entire reason we built Published Monthly at Zen Media.) → Microsoft news

5. MIT Sloan: AI is widening the gap, not closing it.

What happened: MIT Sloan Management Review published “How AI Helps the Best and Hurts the Rest” on April 20. The study reverses earlier findings that AI levels the playing field. Among entrepreneurs facing ambiguous problems, AI made the top performers significantly stronger and made everyone else worse.

What to do: Audit your team’s AI proficiency this week. Identify your top 20%. That is where to invest training time, tooling budget, and air cover. The other 80% needs a structured upskilling program, not access.

Why it matters: This is the screenshot-test stat for the entire quarter. The “AI is democratizing” thesis just collapsed. Your top performers pull further ahead with AI. Your bottom performers fall further behind. The gap is now a permanent line item on your org chart. → MIT Sloan Review

📊 This week’s RECEIPTS, REPORT DROP, and PLATFORM MOVES are below the fold for paid subscribers →

Read more

I had a conversation last week with a CMO friend of mine (for real!) who lost a six-figure deal. The buyer told her: “We asked ChatGPT for the top three vendors in your space. You weren’t one of them.” They are a great company and have been around for TWENTY years. The potential client even loaded the proposal they sent them into ChatGPT and asked if they should work with them.

This is telling of what’s to come.

That’s what this week’s PR@ctical is all about: who AI cites, where the ad dollars are moving, how fast new tools can displace incumbents, and more…

AI is becoming a new layer of “what gets people in the door.”

One thing before we get into it. I started publishing a daily AI briefing on LinkedIn, and now they’re all housed on asksarah.ai/briefings with an RSS feed. Five to seven signals every Monday through Saturday with what to do about each one. If you want it showing up automatically:

1. Feedly: click “Follow Sources,” paste the feed URL, pick a folder. Done.

2. Slack: go to any channel, type /feed subscribe [feed URL], hit enter. Every briefing posts itself.

3. Your website: use RSS.app or Elfsight, paste the URL, copy the widget code, drop it on your blog.

4. Push notification: connect the feed to Zapier, trigger on “new RSS item,” send yourself an email or notification.

OK. Let’s get into it.

🚀 BUILDING & LAUNCHING

Daily AI Briefings: now on asksarah.ai with RSS Five AI signals every weekday morning, each one tied to what you should actually do about it. These have been running on LinkedIn for a few weeks and the response has been wild. Now they’re archived with a full RSS feed so you never miss one. → asksarah.ai/briefings

SourceVault: free AI newsroom research desk I built this for journalists, producers, and hosts. See exactly what media are looking for, who shows up as experts, reverse-engineer your pitches. I think every PR class should be using it. Here’s what an output looks like. → asksarah.ai/tools

⚡ THE BRIEF

1. LinkedIn Is the #1 Most-Cited Source in AI Search for B2B. Here’s Exactly How to Write for It.

SEMrush analyzed 325,000 prompts across ChatGPT, Google AI Mode, and Perplexity and found 89,000 LinkedIn URLs cited in AI-generated answers. For professional and B2B queries, LinkedIn ranks above Wikipedia and Reddit. (SEMrush, April 2026)

The play is NOT to post more. It’s to post differently. The data is specific about what gets picked up and what doesn’t.

What AI actually cites: 95% of citations come from original posts, not reshares. 54-64% of cited content is educational or knowledge-sharing. The median cited post has just 15-25 reactions and no more than one comment. AI does not reward popularity. It rewards relevance and specificity.

The format split matters: ChatGPT and Google AI Mode cite individual member posts 59% of the time. Perplexity flips it: 59% of its citations go to Company Pages. So your personal profile feeds ChatGPT and Google. Your company page feeds Perplexity. You need both.

What gets cited vs. what doesn’t: “Here’s what I learned about content marketing after growing traffic 6,000%” gets cited. “I’m thrilled to announce we raised funding” does not. The first one to three lines carry the claim that gets summarized and quoted.

Your action plan this week:

→ Pick one question your buyers ask before they buy. Not a general topic. A specific question. “How do I evaluate AI visibility vendors?” or “What should a B2B content strategy include in 2026?” Write a LinkedIn post that answers it directly in the first two lines, then supports it with your experience.

→ Use consistent terminology. If you want to be known for “AI visibility,” don’t alternate between “AI search optimization,” “generative engine strategy,” and “LLM marketing.” AI models associate you with whatever language you repeat. Pick your terms and use them every time.

→ Post 5+ times per month. The SEMrush data shows frequency of publication matters more than follower count. A structured cadence of original posts by subject matter experts generates compounding visibility over time.

→ Write one long-form LinkedIn article this month (500-2,000 words). Articles account for 50-66% of all cited LinkedIn content. Feed posts account for 15-28%. If you’re only posting short-form, you’re leaving the highest-citation format on the table.

🔗 SEMrush: LinkedIn AI Visibility Study · LinkedIn: Optimizing Post Content for AI

2. ChatGPT Is Now an Ad Platform. Here’s the Budget Math and What Early Advertisers Are Learning.

OpenAI’s self-serve ads manager is live. The minimum buy dropped from $250K to $50K. Over 600 advertisers are running campaigns. The pilot crossed $100M in annualized revenue in six weeks. (Search Engine Land, March 2026)

This matters because the audience is different from every other ad platform. When someone is talking to ChatGPT, they’re mid-thought. They’re researching a decision. They’re asking “What should I consider when buying X?” That’s bottom-of-funnel intent wrapped in a conversation, and early data from B2B financial services advertisers shows a 1.5x conversion rate versus Google Search. (OpenAI, 2026)

The economics right now:

• CPMs: $18-$65 depending on category. Technology and finance command the top end. Lifestyle and entertainment sit at the low end. Headline rate is ~$60.

• Ad-eligible surface: 900 million weekly active users. 95% on free or Go tiers.

• OpenAI expects $2.5B in ad revenue by end of 2026.

What early advertisers are learning:

The brands winning are the ones that match the conversational context, not the ones running traditional display creative. When someone asks “What should I consider when buying a standing desk for a small apartment?” that’s purchase intent even though it doesn’t look like a search query. E-commerce, SaaS, education, healthcare, and financial services are the five industries seeing the strongest early results. (AdventurePPC, 2026)

Your action plan this week:

→ Pull your current Google and Meta CPMs for Q1. Put them side by side with ChatGPT’s $18-$65 range. If your Google CPMs are above $60 in a tech or finance category, ChatGPT might actually be cheaper with higher intent.

→ Structure your first campaign as a learning investment, not a performance play. OpenAI recommends testing different conversation contexts, varying educational vs. promotional tones, and tracking indirect signals: branded search volume increases, direct traffic changes, brand lift. Don’t optimize for immediate ROAS. Optimize for learning what works in conversational context.

🔗 OpenAI: Testing Ads in ChatGPT · Search Engine Land: ChatGPT Ad Revenue

3. Claude Design Launched Friday. Figma Dropped 7%. Here’s What This Means for Your Team’s Bottleneck.

Anthropic released Claude Design on April 17, powered by the new Opus 4.7 model. Describe what you need in a conversation and it builds prototypes, pitch decks, one-pagers, website drafts, and social cards. It reads your design files and applies your brand system automatically. Available to Pro, Max, Team, and Enterprise subscribers at no additional cost. (9to5Mac, April 2026)

The market reaction is pretty darn telling. Figma dropped 7.28% to $18.84 by close. Adobe fell 2.7%. Wix dropped 4.7%. GoDaddy declined 3%. Why? Because investors can see we are entering a world where non-designers can produce professional-quality visual work from a text conversation. (Yahoo Finance, April 2026)

Sarah’s take: My hope is that this ends the “design bottleneck.” Claude Design doesn’t replace your design team. It eliminates the 80% of requests that shouldn’t be in the design queue in the first place. The “can someone make this a slide deck by Thursday” work just got automated.

Your action plan this week:

→ Identify your team’s top 3 design bottleneck requests from the last month. The slide deck for the board meeting. The one-pager for the sales team. The social cards for the campaign launch. Pick the most common one.

→ Run a timed test. Open Claude Design, upload your brand guidelines (PDF or design file), and ask it to build that deliverable. Time yourself. If you get to 80% quality in under 10 minutes, bring that number to your next ops meeting.

→ Do the math on design queue time. If your average design request takes 3-5 business days in queue and Claude Design gets you 80% in 10 minutes, what does that free up? Not in theory. In actual hours your team gets back this quarter.

🔗 Gizmodo: Claude Design Launch · Anthropic: Claude Opus 4.7

4. Adobe Summit Opened in Vegas. Every Session Is About AI Agents Running Your Marketing.

Adobe Summit 2026 kicked off this weekend with 200+ sessions on one theme: AI agents that manage campaigns, personalize journeys, and create content without constant human direction. GenStudio, Journey Optimizer, and Real-Time CDP are all getting agent features. Nearly double the AI exhibitors compared to last year. (CXToday, April 2026)

Two agents worth paying attention to. The content production agent generates brand-compliant assets from a creative brief. The site optimization agent runs continuously, detecting and fixing engagement issues without a human in the loop. Traditional batch processing is being replaced by real-time: audiences update continuously and messages fire the moment a behavioral trigger occurs.

Sarah’s take: I’ve been watching Adobe Summits for years (both as an Adobe Influencer and attendee). This is the first one where the entire agenda is one topic. Adobe is betting that by this time next year, the companies using agents for content production, A/B testing, and audience segmentation will have a structural advantage over the ones still running those workflows manually. Based on what I’m seeing with our clients at Zen Media, I think they’re right.

Your action plan this week:

→ Audit your team’s top 5 repetitive marketing workflows. Email A/B test setup, audience segmentation, content adaptation across channels, campaign reporting, social scheduling. Rank them by hours spent per week.

→ Pick the one with the highest hours and lowest strategic value. That’s your first agent candidate. You don’t need to buy anything yet. You need to know which workflow you’d hand off first so that when your Adobe rep calls (and they will), you have a specific answer, not a vague “we’re interested.”

→ Ask your marketing ops team one question: “If an AI agent could run one workflow end-to-end starting next quarter, which one would free up the most time for strategic work?” Write down their answer. That’s your Q3 pilot.

🔗 CXToday: Adobe Summit 2026 · Stensul: Adobe Summit Guide

5. The $200K+ Job That Didn’t Exist Two Years Ago. And What the Hiring Pattern Tells You.

On Thursdays I’ve started sharing new roles in our industry that most people don’t know about yet. Credit to Noah Greenberg at Stacker, who rounds up non-traditional hiring for content and journalism roles. I wanted to do the same for AI visibility.

I went through LinkedIn, Indeed, Glassdoor, and SEOjobs and pulled 18 open roles that include AEO, GEO, or AI search visibility in the job title or core responsibilities. Here are the ones that should get your attention:

• Stripe: AEO & GEO Marketing Manager / $138,500-$207,800 / Remote US

• ADT: Director of Discoverability, AEO/SEO / Est. $153,000-$286,000 / Irving, TX

• Pfizer: Manager, SEO and AI Discoverability / $106,000-$172,000 / New York

• Profound: Senior AI Strategist / $150,000-$200,000 / NYC, SF

• Synchrony: AVP, Technical SEO & GEO / $90,000-$155,000 / Multiple US

• Eliassen Group: Generative Engine Optimization Lead / $156,000-$166,000 / Remote

• Experian: AEO & SEO Manager / $90,000-$130,000 / Remote US

• Capital One: Sr. Digital Marketing Associate, AEO/GEO/AI Search Lead / Nottingham, UK

Plus 10 more from Odoo, Apryse, Hawksford, Vector Security, AccuraCast, and others.

Sarah’s take: The list is interesting. The pattern is more interesting. Stripe, Pfizer, ADT, Synchrony, Experian, Capital One: these are not early-adopter startups experimenting with a buzzword. These are enterprise companies building permanent teams around a discipline that barely had a name in 2024. When Pfizer creates a “Manager, SEO and AI Discoverability” role, that’s a signal that AI visibility is now a line item in the pharma marketing budget. When Stripe builds this into their marketing org, it’s because they’ve seen the data on where their buyers are shortlisting vendors.

The skill set across all 18 listings follows the same pattern: PR + SEO + structured content + understanding how AI models decide who to cite. That combination is rare right now. It won’t be in 18 months. The salary ceiling on these roles is over $200K because the supply of people who can do this work is tiny relative to the demand.

Your action plan this week:

→ If you’re a CEO or CMO: open one of these job descriptions and read the requirements section. Then ask yourself: does anyone on my team have this skill set? If the answer is no, you have a gap that your competitors at Stripe and Pfizer are already filling.

→ If you’re a practitioner: look at the skills listed across these roles and identify your biggest gap. Is it structured content? AI citation mechanics? SEO technical fundamentals? Pick one and spend 30 minutes this week learning the basics. The career path is real and the window to build expertise early is still open.

→ If you’re hiring: don’t wait for the perfect candidate with all four skills. Hire for two and train for two. The people who have PR instincts plus willingness to learn structured content and AI citation mechanics will outperform the people who only know SEO and are trying to learn the communications side.

📊 This week’s RECEIPTS, REPORT DROP, and PLATFORM MOVES are below the fold for paid subscribers →

Read more

Hi Team PR@ctical.

Last week, someone at Meta built an internal leaderboard called “Claudeonomics.” It ranked 85,000 employees by how many AI tokens they consumed. The top users earned titles like “Token Legend” and “Session Immortal.” In 30 days, Meta employees burned through 60 trillion tokens (The Information, April 2026).

Then it got shut down. I won’t go into that, but it got me thinking…

If you don’t know what a token is or the infrastructure that powers AI, you won’t have everything you need to use it at scale (not to mention what you can do with it).

A token is the smallest unit an AI model processes. One token is roughly 4 characters or about three-quarters of a word. When you type a prompt into ChatGPT or Claude, your input gets broken into tokens. The AI’s response is also measured in tokens. Every AI interaction has a token cost: you’re charged for what you send in and what the model sends back.

Why does this matter? Because tokens are how companies price AI. They’re how enterprises measure usage. Jensen Huang tells Nvidia engineers they should consume tokens worth 50% of their annual salary (Tom’s Hardware, 2025), he’s saying: use AI aggressively or fall behind.

This week, I’m also launching an AI Notices Certification. Coming shortly! It’s a 60-slide video masterclass with a workbook, templates, and exercises that teaches you how to write the kind of content AI systems actually cite.

And it’s step one. I’m also building the AI Authority Program: a full certification ladder for PR and comms professionals who want to upskill and be ready for the future with AI.

Big week. Let’s get into it.

🚀 BUILDING & LAUNCHING

AI Notices Certification: The Complete Masterclass | $197 | Coming Shortly The only training that teaches you how to write content AI systems actually cite. 60-slide video masterclass with full audio narration (~40 minutes), companion workbook with fill-in templates and checklists, printable slide deck for reference.

What you walk away with:

• A wire draft template you can use on day one

• The AI Notices Submission Table template (the single source of truth for every engagement)

• A complete SOAR self-audit checklist

• JSON-LD schema template ready to customize

• A citation tracking log to prove ROI to clients

• 3 practice exercises with full answer keys

Built for PR and comms professionals adding AI visibility as a service, in-house teams building their own strategy, and consultants adding a new revenue line. Step one of the AI Authority Program.

🔗 asksarah.ai

Coming Soon: Claude Training Series Paid trainings on setting up and using Claude for PR workflows: chats, projects, scheduled tasks, and skills. I broke down how I built Scout (my AI chief of staff inside Claude Cowork) on LinkedIn this week. Start with a Claude Project, write custom instructions like you’re onboarding a human, build skills for your specific workflows, connect your tools, and let it learn you over time. If you can describe how you want something done, you can build a skill for it. Full trainings dropping soon.

⚡ THE BRIEF

1. Meta’s Employees Were Racing on an AI Leaderboard Called “Claudeonomics.” Then It Got Shut Down.

An employee-built dashboard called “Claudeonomics” (named after Anthropic’s Claude) tracked the top 250 AI power users across Meta’s 85,000 employees. Titles included “Token Legend,” “Session Immortal,” “Cache Wizard,” and “Model Connoisseur.” The numbers: 60+ trillion tokens consumed in 30 days. At public pricing, that’s an estimated $9 billion in token costs (The Information, April 2026).

The kicker: neither CEO Mark Zuckerberg nor CTO Andrew Bosworth ranked in the top 250 (Fortune, April 2026). Employees were reportedly running idle AI models just to inflate their rankings.

Meta shut it down within 48 hours after the data leaked externally. But the leaderboard was a symptom, not the cause. Meta has now officially tied employee performance reviews to “AI-driven impact” (The Information, April 2026). CTO Bosworth’s position: if a top engineer spends $500K on AI tokens and gets 10x productivity, “It’s a no-brainer. Keep doing it. There is no cap.”

What to do: This story is a preview of what’s coming to every enterprise. If your clients are measuring AI adoption, make sure they’re measuring outcomes, not consumption. Volume without results is a leaderboard that gets shut down.

🔗 The Information | Fortune

2. Half of U.S. Workers Still Won’t Touch AI. The Number One Reason: “I Like Doing It My Way.”

Gallup surveyed 23,717 employed adults in February 2026. Half of American workers use AI once a year or not at all. Among those who have access and still refuse, 46% say they prefer doing work the way they always have. 4 in 10 cite ethical concerns or data privacy worries. Meanwhile, more workers than ever say they’re worried AI will replace them (Gallup/ABC News, April 2026).

For context: Cision’s “Inside PR 2026” report shows 91% of PR professionals now use AI, with 73% using it for ideation and 68% for writing (Cision, April 2026). So our industry is ahead of the curve. But the broader workforce gap is massive, and that gap affects your clients.

What to do: Training and culture beat software rollouts every time. (This is exactly why I built the certification.)

🔗 Gallup | Cision

3. 74% of AI’s Economic Gains Are Going to Just 20% of Companies. Everyone Else Is Stuck in Pilot Mode.

PwC surveyed 1,217 senior executives across 25 sectors (April 2026). Three-quarters of AI’s total economic value is being captured by one-fifth of organizations. The companies pulling ahead are 2.6x more likely to use AI to reinvent their entire business model. Companies with Responsible AI frameworks are 3x more likely to report real financial returns (PwC, April 2026).

Meanwhile, OutSystems surveyed 1,900 IT leaders: 96% are already using AI agents, but 94% say sprawl is increasing complexity and security risk. Only 12% have a centralized platform to manage it all (OutSystems, April 2026). Writer’s enterprise survey found 54% of C-suite executives say AI adoption is “tearing their company apart” (Writer, 2026).

What to do: Everyone deployed agents. Almost nobody built the governance layer first. For comms teams, this is an opportunity: help clients tell the story of responsible deployment, not just deployment.

🔗 PwC | OutSystems | Writer

4. Big Tech Committed $700+ Billion to AI in 2026. Earnings Season Starts This Week.

Amazon: $200 billion. Alphabet: $175-185 billion. Meta: up to $135 billion. Microsoft: ~$150 billion. The hyperscalers have committed $700+ billion combined to AI infrastructure this year (CNBC, February 2026). Goldman Sachs estimates AI investment will account for 40% of S&P 500 earnings growth in 2026 (Goldman Sachs, 2026).

Q1 earnings season kicks off mid-April. Alphabet reports April 23. Amazon shares already dropped 11% after revealing the $200 billion forecast. Pantheon Macroeconomics reports that without AI spending, U.S. corporate capex would be negative (Fortune, February 2026). The entire capital investment story in America right now is an AI story.

What to do: When your CEO or client asks whether the AI budget is justified, point here. Use these numbers to frame internal investment conversations and budget approvals.

🔗 CNBC | Goldman Sachs | Fortune

5. Six U.S. States Passed AI Laws in the Last Two Weeks. More Are Coming.

• Nebraska (LB 525): conversational AI interacting with minors must disclose it’s not human. Bans manipulative reward systems and AI simulating emotional dependence. Effective July 1, 2027.

• Maine (LD 2082): no therapy through AI without a professional license. Sent to governor April 10.

• Maryland (HB 895): bans dynamic pricing in grocery stores using surveillance data. Penalties start at $10,000.

• Tennessee (SB 1580): AI systems cannot represent themselves as mental health professionals. Effective July 1, 2026.

• Oregon (SB 1546): chatbot operators must implement protections for children. Effective January 1, 2027.

• Idaho (S 1297 + SB 1227): chatbot safety + statewide education AI framework. Explicitly states AI tools cannot replace human teachers.

The DOJ established an AI Litigation Task Force in January 2026 to challenge state AI laws deemed unconstitutional (Troutman Privacy, April 2026).

What to do: If your company or clients use chatbots, AI assistants, or any customer-facing AI, a state-by-state compliance map is overdue. For comms teams: draft disclosure language now, before you’re required to.

🔗 Troutman Privacy

📊 This week’s RECEIPTS, REPORT DROP, and PLATFORM MOVES are below the fold for paid subscribers →

Read more

Hi Team PR@ctical.

Quick housekeeping before the data: we lost OpenClaw this week.

🪦 RIP.

If you don’t know what it was, that’s fine. If you do, you know why this matters. I did absolutely wild things with it. Built an AI Chief of Staff who named herself Kai. Shipped 27 new products. And so much more I don’t even have the space to share.

Anthropic shut the door on subscription access on April 4 because the math wasn’t mathing on their side.

Biggest takeaway?

I saved every output, had Antigravity logging the whole run, and Claude Code and Cowork were already standing by. I spun up "Scout" inside Claude, and he's great. Prepare for bright, shiny new (and amazing) tools, and create infrastructure for when they go away and something new takes their place.

While that door closes, here’s a door I want you to actually walk through: my dear friend Shama Hyder’s REWIRED Mastermind. Jump. On. This.

Shama is one of the few people building an AI mastermind that’s actually about how to think, not just which tool to buy. This is the room to be in.

And now… the theme of today’s edition:

That means at a basic level you need to understand that LLMs, AI search, and answer engines are changing consumer behavior, and you need to understand how that impacts your role.

1. Spend 30 minutes a day in the tools. Not reading about them. In them. Run your own brand through ChatGPT, Perplexity, Claude, and Gemini. Read what they say back. That is your new earned media report.

2. Learn one workflow, not one tool. Pick a thing you do every week (media list building, statement drafting, briefing memos) and rebuild it with AI in the loop. Workflow architects win. Tool collectors lose.

3. Translate it for your team. If you can’t explain to a junior what an LLM is and why it cites what it cites, you are not yet AI literate. You are AI adjacent.

The DOL just made this a federal workforce standard. Muck Rack says 76% of us are using AI and 12% understand agents. The gap between use and understanding is the entire story right now.

🚀 BUILDING & LAUNCHING

GEO GPT — AI Visibility Engine Plug in a brand, a topic, or a prompt set. GEO GPT tells you where you’re showing up in AI answers, where you’re not, and what to fix first. Built for the audit you should be running this week, not next quarter.

The Knowledge Panel & AI Visibility Playbook (25 min read) The full playbook for getting your brand into AI answers, knowledge panels, and the citation layer LLMs actually pull from. If you’re going to read one thing this week, read this.

⚡ THE BRIEF

1. The DOL made AI literacy a federal workforce standard. What happened: The Department of Labor released its AI Literacy Framework in February, defining AI literacy as a foundational capability all workers should have. It is now the federal baseline for what “AI literate” actually means. What to do: Pull the framework. Use it as the rubric to audit your own team’s skills, not your tool stack. If your AI training plan is “we bought ChatGPT Enterprise,” you do not have a plan. Why it matters: When the federal government codifies a skill as foundational, RFPs, hiring criteria, and procurement language follow. Comms teams that can document literacy will win business that “we use AI” teams will not. 🔗 DOL AI Literacy Framework

2. 76% of PR pros use AI. 12% understand agents. What happened: Muck Rack’s State of AI in PR 2026 dropped. 76% of PR pros use generative AI, 75% pay for at least one tool (up from 57% last year), and 98% always or often edit what AI gives them. But only 12% have adopted AI agents into their workflow even though 60% have heard of them. What to do: Run a 30-day agent pilot on one workflow. Just one. Media monitoring, briefing memos, or executive prep are the easiest entry points. Why it matters: The teams that figure out agentic workflows in 2026 are going to bill the same and deliver 3x. The teams that don’t will get repriced by their clients. 🔗 Muck Rack State of AI in PR 2026

3. 91% of PR pros use AI. Storytelling is still the #1 in-demand skill. What happened: Cision’s Inside PR 2026 found 91% of PR pros use generative AI in some part of their workflow. The most in-demand skill for the year? Storytelling, at 59%. Followed by media relations and strategic planning. What to do: Stop treating AI literacy and storytelling as opposing forces. The strongest comms pros in 2026 will be the ones who can do both, in the same draft, on the same deadline. Why it matters: AI is not replacing the craft. It is raising the floor. If your only differentiator was “I can write fast,” your floor just moved. 🔗 Cision Inside PR 2026

4. 82% of journalists are now using AI. They are using it on you. What happened: Muck Rack’s State of Journalism 2026 found 82% of journalists use at least one AI tool (up from 77%), and 26% cite “unchecked AI” as a top industry concern. What to do: Audit every pitch you send for the next 30 days against this question: “If a journalist runs this through an LLM, what does it say back?” If your boilerplate is generic, you are getting deprioritized in real time. Why it matters: Your media list is already in the loop. Your pitches are being summarized, fact-checked, and cross-referenced before a human ever reads them. (Shameless plug: this is exactly what we built Published Monthly at Zen to solve.) 🔗 Muck Rack State of Journalism 2026

5. Comms leaders are over-indexing on tools and under-indexing on judgment. What happened: PR Daily called this the most dangerous AI gap of 2026: comms teams are buying tools faster than they are building the judgment to use them. What to do: Before you buy anything else, write a one-page “what we will not use AI for” doc. It is the fastest way to find out where your team’s judgment actually lives. Why it matters: Tool sprawl is the new reputational risk. The teams that name the line get to defend it. 🔗 PR Daily: closing the most dangerous AI gap

📊 This week's RECEIPTS, REPORT DROP, and PLATFORM MOVES are below the fold for paid subscribers →

Read more