I published my first article on best Facebook ad examples in 2018.

Since then, it’s become my habit to capture any great ad creative I spot on Instagram or Meta Ads Library. I save them in a Figma doc (image ads) and Google Drive (videos).

My personal gallery of Meta ad ideas holds well over 1,000 creatives.

Whenever I work on a new brand’s paid ad campaign, I return to this gallery to collect ideas. I also check out the competitors’ Ad Libraries (which is how my own collection expands).

The best thing about having a mixed library of B2C and B2B brand ads is that it helps you come up with out-of-the-box ideas. My go-to trick is to use consumer brand ads as inspiration for my B2B tech clients’ creatives, and vice versa.

For example, a recent top-performing ad I created for Roleo (B2B recruitment) was a fun adaption of an AG1 ad (B2C health supplement).

What works in Meta ads in 2026?

I’ve noticed 3 key changes in the Meta ad creatives discourse over the past 12 months:

1. Self-proclaimed Meta ad experts on LinkedIn post about using AI tools to generate “100 creatives in 30 minutes.”

If you’ve ever tried out one of their prompts, you know that’s a lie.

High-performance ad creatives are still created by human designers, mostly in Figma (static ads) and Adobe After Effects (video ads).

For example, SupersideHQ’s video ad was likely done in After Effects. It combines high-quality video clips, product UI, and text animations.

2. It’s become much easier to create OK video ads.

Meta recommends running a mix of static and video ads, with more focus on UGC (user-generated content style) videos.

The most beautiful and complex video ads still require the Adobe After Effects skills.

But you can design pretty decent-looking videos that perform with CapCut and Instagram Edits.

Below is an example of a recent video ad I created in CapCut.

3. Competition for people’s attention has never been tougher.

Whenever it gets easier to produce something, there’ll be more of it.

Everyone’s tired of AI slop, and even the most excellent original content suffers from this. This applies to newsletters and blogs, social media posts, and your paid ad creatives.

To stand out in 2026, start from defining your ICP (ideal customer profile). Who are these people in real life? What are the aesthetics that catch their attention? What’s the best value prop to promote? What’s fun and engaging for them?

Wise's Ad Library is one I check again and again. They always lead with relevant USPs in bold headlines, build comparisons, and address specific jobs to be done.

How to assess a brand’s Meta Ad Library?

There are millions of ads in the Meta Ad Library.

How do you know which ones actually perform, i.e. worth copying?

The short answer is: experience.

When selecting the new ad examples for this article, I asked myself what I’d be interested in copying myself for my clients’ projects.

The creatives that stood out follow five best practices that have stood the test of time. All were already relevant back in 2018.

1. Communicate a strong value proposition inside the ad creative.

2. Use high-quality images and videos. Avoid cheap stock material.

3. Experiment with bright colours and high contrast.

4. Add a clear call-to-action in the ad visual.

5. Ads with numbers (e.g. discount, % of improvement) and recognisable logos perform best 90% of the time.

Top-performing Meta ads in 2026

Alright, let’s get to it!

Scroll on for my personal eight favourites from the 2026 batch.

Some are in portrait format, which means they run both in Feed and Story placements.

Due to newsletter length limitations, you can find all 240+ Facebook and Instagram ad examples on my blog.

See 240+ Facebook ad examples

1. Lovable

Why I think it’s great: Strong number-led value prop (save 10k), bold and instantly recognisable brand colours, a lot of text in ad image (nobody reads the text description).

2. Claude

Why I think it’s great: Claude’s (Anthropic) Meta Ad Library has 550 active ads. Why? Because they’ve created specific ads for all their ICPs: developers, founders, designers, etc.

This one’s for marketers. Clean design, familiar integration logos, carousel cards that show real work benefits.

3. Healf

Why I think it’s great: Unconventional ad format that stands out in Instagram Feed, collaboration with a well-known health influencer, number-led “savings” value proposition.

4. Canva

Why I think it’s great: Collaboration with an expert explaining how to use the tool, high-quality UGC video recording, product UI shown in the ad, teaching while promoting the product.

5. Vinted

Why I think it’s great: Collaboration with a major local Gen-Z influencer to show how easy it is to sell on Vinted. On the other hand, the ad invites Aimee France’s followers to sign up to see her Vinted store.

6. Notion

Why I think it’s great: Recognisable brand style, ads with white/light background pop out in the otherwise colourful Instagram Feed, super valuable discount offered, clear CTA inside ad image.

7. Revolut

Why I think it’s great: Question that invites the viewer to think and engage, beautiful design that makes you want one of the cards, portrait-format creative that can be adjusted for Feed placements by Meta’s algorithms.

8. AG1 by Athletic Greens

Why I think it’s great: I like how they’ve made a photo-based ad so much more than that. The bright green headline, overlays with supplement information… And managing to convey so much information by keeping the overall look uncluttered and on-brand.

Thx for reading!

Don’t forget to check out the full blog article with more examples.

For more paid ads advice and inspiration, see these three articles:

• Fix your landing page header for 30% conversion uplift: 6x A/B Tests to Digivolve Your Web Header into 2026

• Get your ad formats right: Facebook Post & Ad Sizes in 2026: Foolproof Guide

• Double down on inspiration: 105+ LinkedIn Ad Examples: 2025 Fast-Growth Brands

Have a good summer!

And keep thinking w/ your own head.

Karola

If you find it useful, forward it to a tech founder or marketer in need of inspiration.

The best marketers don’t excel in one single channel.

It’s not about being a t-shaped marketer, either.

And it’s also not about good taste, as the business media and influencers have been touting for the past six months.

What makes a good marketer is the ability to learn and adapt.

Let me explain.

Between 2015 and 2026, Meta ads became 5-10x more expensive, and AI killed 90% of organic website traffic from Google.

But it doesn’t mean that paid social ads or content marketing no longer work.

Marketers just need to approach these channels differently.

Facebook and Instagram ads now require different ad formats, and higher-quality designs that communicate more, faster.

Content marketing, when done on a website, still requires a high domain authority and fast loading speeds. Your articles still need to rank in top Google 1-5 spots to be picked up by AI tools and listed in query replies. You just need to track the results differently.

But the truly efficient content marketing now happens in newsletters and social media posts.

This necessity to adapt (to new technologies and trends) also applies to early-stage brand, marketing, and growth.

The principles are 100% the same in 2026 as they were back in 2015.

So w/o further ado, here they are.

10 principles of brand, marketing, and growth for early-stage companies

After helping 50+ European tech startups with early-stage growth, here are the key learnings that I continuously bring to all new projects.

Brand:

1. Invest in strong (enough) brand early.

2. Understand your ICP (ideal customer profile).

3. Create what your ICP wants.

Marketing:

4. Focus on 2-3 marketing channels.

5. Copy what works for other brands.

6. Prioritise quality over quantity.

7. Multiply your marketing budget 5-10x.

Growth:

8. Build growth loops. 1 user » 1.5-2.5 users.

9. Run 2 experiments/month, minimum.

10. Don’t worry about monetisation (just yet).

Bonus: Hire top experts & excited beginners.

1. Invest in strong (enough) brand early

In my humble opinion, the biggest mistake that early-stage startups make is not investing in a strong brand.

This is especially true in Eastern Europe and the Baltics.

“It’s the product that people care about,” the practical mind says.

And I do agree. The product does have to be great. However, the same applies to your brand.

Why?

Imagine meeting a new person. You either like them or not.

It’s exactly what happens when you go to a company’s website (or App Store page) for the first time.

You either like the brand or you don’t.

As an early-stage company, you only have one chance to make a good first impression with each potential customer.

People don't return to brands they didn’t like in the first place. (Either because of a cheap-looking brand or bad product experience.)

If I could do one thing differently in early-stage Taxify (now Bolt)…

In hindsight, I wish we’d invested in a high-quality brand that looks trustworthy to people in Western Europe, too.

Bolt now has a world-class brand.

But I can’t help but think that we’d grown much faster in London and Paris, had we launched there w/ a higher-quality brand.

2. Understand your ICP (ideal customer profile)

In order to build a successful product and brand, you need to understand your target audience.

Who is your ICP – ideal customer profile? Is there a clear understanding about it across the organisation?

The ICP is focused on your most valuable customers and prospects who are also most likely to buy.

When you google customer persona development guides, they often look like personal CVs.

This is not a sufficient way to define a customer persona.

You want to have a deeper understanding of their behavioural patterns, lifestyle, everyday struggles and needs, touchpoint marketing channels, etc.

Below is 1/10 of the ideal customer persona mapping sheet I use on early-stage brands.

3. Create what your ICP really wants

This is true for both your product features and marketing assets.

The not-so-nice truth about working w/ early-stage companies is watching many of them creating a product that nobody wants or needs.

On the marketing side, once you know your ICPs (ideal customer personas), you can develop value propositions that:

1. … are actually useful and relevant to them.

2. … you can truly deliver on.

3. … differentiate you from the competition.

Both your product and marketing should solve your ICP’s pain points.

And by pain points I mean a pain so great they’re willing to break the status quo and give your product a go.

For example, Wise knows their ICP’s main goal is to save money on transfers.

They’ve added a comparison of major banks right on their home page.

4. Focus on 2-3 marketing channels

… instead of doing 10+ channels badly.

I often have to negotiate w/ CEOs who’ve read some LinkedIn post about a new marketing channel X and think we need to try it.

“But why do you think it’s more efficient than our current channels that deliver positive ROAS?”

Silence.

Especially in early stage, when the founding team members’ time is your most valuable asset, it’s practical to focus on fewer marketing channels.

Choose 2-3 marketing channels that reach the maximum number of your ICPs.

In B2C, it could mean organic social media, and paid social (Meta) ads. Or an OOH marketing campaign. Or PR stunts. Or Reddit. Or building a content system that helps your product get referred by AI tools.

In B2B, LinkedIn (organic + paid), content marketing optimised for AI agents, and marketing-supported sales outreach work well. Partnerships and affiliate programmes also, when executed well, return a high ROI.

As a performance marketer who loves to experiment w/ ad designs, I usually start from paid social (Meta + LinkedIn) to launch user acquisition fast. Later, I add a few additional channels, depending on the ICP and product.

5. Copy what works for other brands

There’s no such thing as innovation or uniqueness in contemporary marketing. It’s all tried and tested tactics and channels.

What matters is how you approach each channel.

Research what’s working for other brands.

Don’t copy your competitors, but other brands who target your ICPs.

The easiest way is to check their website, social media channels, and paid ad libraries (Meta and LinkedIn).

See where they invest most.

Is it content, creators, partnerships, paid ads, organic social, live events, webinars, etc.? Is their sales team active on LinkedIn?

For example, a quick scroll through Clay’s LinkedIn ad library tells me that they run lots of webinars w/ partners and creators, and publish in-depth industry reports.

6. Prioritise quality over quantity

“This marketing channel doesn’t work for us.”

Sounds familiar?

Sometimes, it’s fair to scrap a marketing channel (or experiment) and admit defeat.

More often, marketers say “it’s not working” after testing a channel for 72 hours.

Don’t scrap a marketing channel because of your own scrappy work.

Give it at least 30 days of hard work.

Another problem I often see during marketing audits is the low quality of marketing assets.

I’ve seen dozens of company blogs that have 50+ articles. But none of them is actually good… Nor ranks on the 1st page of Google search results.

By “actually good” I mean the best article that exists on the given subject.

For example, when compiling my free library of 95 marketing audit questions, I didn’t leave it at 15. I went all the way, creating the most extensive resource on the subject.

It’s the #1 search result for the Google query of “marketing audit questions.” And the first resource quoted in the AI summary.

(I tested in an incognito window to remove search bias. Do it whenever checking your own keyword rankings.)

One thing that’s changed between 2015 and 2026 is the level of quality people expect from brands.

Nobody needs more low-quality slop. There’s already too much in the world.

On the other hand, it’s so easy to feel like you have to do something. Because, like other human beings, you encounter 100x more cool stuff daily than you used to five years ago.

Don’t fall into the good marketing anxiety trap.

Instead of despairing “there’s so much good stuff out there, how can I even compete?” take a step back from the noise.

Zone out. Book a 4-hour slot to think. And come up w/ something that’s valuable to your ICP that you can deliver.

That something probably isn’t something you can create in 10 minutes w/ an AI tool. If it is, it already exists.

It’s the slow, focused, demanding work that translates into high-quality marketing assets and campaigns.

I’m not saying “original” on purpose. I don’t believe that in 2026, originality matters as much as quality and craft.

Spend more time on creating quality marketing assets (articles, landing pages, ad creatives, PR stunts, newsletters, etc.) rather than launching semi-good ones.

7. Multiply your marketing budget 5-10x

Hot take, I know.

Isn’t an early-stage startup supposed to be… FRUGAL?

Well, yeah.

But the thing it needs to be frugal about is not money. It’s time.

Each month that your company fails to grow exponentially, you’re burning your funding on team salaries, office rent, software licenses, etc.

I’ve seen so many startups die slowly while playing it safe.

And so have you.

The whole point of having an investment (vs bootstrapping) is to grow fast.

Early-stage companies are like ice cream. : p

You either go all in or watch it melt through your fingers and stain the floor (your reputation).

Would you rather work w/ a second-time founder (or founding marketer) who…

a) Made bold early bets which didn’t work out, and failed in 12-24 months.

b) Kept the company barely functioning for 10 years, then folded.

I know which I’d prefer.

A lot of the time, the best moat for tech companies is being the first to take the market.

Chances are that you’re underspending on marketing and expansion.

Here’s how I think about early-stage growth investment:

1. Spend cautiously until you’ve found your product market fit, positive marketing ROAS, and functioning growth loops.

• Once you’re growing organically (word of mouth, referrals = people love your product), 5-10x your marketing budget and expand to new markets fast.

For example, here’s Klarna’s expansion timeline.

8. Build growth loops. 1 user » 1.5-2.5 users

All successful tech companies have a growth loop built into their product.

A growth loop means that every new user you bring into your product will invite others.

It also means that if you spend €50 to bring in a new active user via paid ads, the CAC (customer acquisition cost) gets diluted once they bring other people into your product.

The most obvious growth loop is referrals. But there are many more.

Here’s an excellent article by Maja Voje on growth loops, including many examples.

9. Run 2 experiments/month, minimum

The best way to find things that work in the early stage is to experiment.

This applies to both marketing and product bets.

Rule #1 of experimentation: do it regularly and consistently.

In the past months, working on Roleo’s marketing and growth, I’ve run at least one Meta ads experiment every week.

There are weeks when the new ad creatives perform worse than the previous ones.

And then there are weeks when I discover a new creative that brings 2x as many new users at a 50% lower CAC.

Rule #2 of experimentation: prioritise the “highs.”

Often, low-effort ideas get prioritised over high-effort and high-impact.

Stay alert to this trap.

Also, give priority to high-confidence ideas.

You have a solid understanding of your ICPs and their pain points. Trust your gut feeling. When in doubt, validate any new idea by talking to your users (and non-users).

My take on user interviews:

You want to talk to people who use and love your product.

But more importantly, you should talk to your ICPs who are not using your product to understand what you need to build and improve.

10. Don’t worry about monetisation (just yet)

Another controversial view, I’m aware.

“If people aren’t willing to pay for our product, there’s no product-market fit,” you might argue.

Yes, I agree to some extent.

But I’ve seen too many early-stage startups limit their early-stage growth with a bad monetisation model.

By “bad” I mean a a pricing model that stops new customers from trying the product and stalls both supply and demand.

In the early stage, use your funding to grow, not to make back 0.1% of your potential future earnings.

In order to monetise a market, you first need to take it.

For example, when entering a new market, Bolt burned millions on rider discounts and driver bonuses before reversing the profitability chart.

If you think about fast-growth AI companies in 2026, they subsidise computing power, tokens, and software features.

Or they give consumers free agentic products that, in the background, cost them millions of €$£ in computing bills.

These companies will become profitable in 2 or 5 or 10 years.

But first, they focus on fast growth and taking the market. To build the moat that makes it nearly impossible for any new players to enter.

Next steps

Scroll back to the numbered list of 10 early-stage company marketing rules.

Take 10 minutes to reflect on them.

1. Which ones do you already check?

2. Are there any that might help you grow faster?

3. Do you have the right team to implement them?

Speaking of the bonus rule of hiring top experts and excited beginners…

If you need the expert side, I’m open to 1-2 new consulting gigs, starting from September 2026. Learn more here or say hi on LinkedIn.

Have a good summer! And keep thinking w/ your own head.

Karola

Forward this to your friends running paid ads for inspiration & quick wins.

If you're running paid ads, your creatives are the number one thing that decides your success. (By success, I mean ROAS and revenue, not impressions and clicks.)

Number two is something you've barely touched since the last rebrand: your website header.

You get the click. Fine.

But what’s the website conversion into leads and paying customers?

Quick maths: if you spend €10,000/month on paid ads and your signup CAC is €28, improving your website conversion rate by 20% buys you 71 extra signups a month.

That’s €24,000 a year in signups you didn’t have to pay for. Or a 20% bump in new revenue added.

But here's where most marketers get it wrong.

They redesign the entire website to improve the conversion rate.

Nothing below the fold matters. Unless your header gives people a reason to scroll.

Visitors judge your website in 50 milliseconds.

That’s 2.5x faster than a blink of an eye.

It is your website hero that decides whether people will sign up and buy.

In the previous issue of Marketing Fix, I analysed the hero copy of 15 category-defining AI products.

I promised the design half next, so here it is.

Light vs. dark, product UI vs. illustration, typography trends and product demos. And because looking at pretty websites fixes nothing, each finding comes as an actionable test you can run on your own header this week.

So that your website header can digivolve from Patamon… into Angemon.

Six AI-era website header design trends

Here’s a quick overview of six key trends in AI brands’ website headers.

These brands employ some of the best design and marketing minds money can hire. Which makes their headers less "inspiration" and more free R&D.

Everything below is worth a copy-and-run test.

1. Bin the 50/50 layout. Choose a single focus point.

2. Ditch the white default. Pick your own mood.

3. Forget Inter & Helvetica. Go for a human serif.

4. Stop asking for the click. Let visitors try the product.

5. Cut the stock photos. Go bare or baroque.

6. Change the hero weekly. Keep the logo bar sacred.

Let’s take a look at what each of these means in practice, and how you can turn them into A/B tests for your own website.

1. Single focus point » Headline 1

Examples: Whispr Flow, Dovetail, Clay

Back in 2016, website headers balanced headline and image equally.

In 2026, all good marketers know that 80% of visitors are on their smartphones. Just check your paid ads traffic and placements.

This means that you’ll have to choose one thing to show first: image or message.

Nearly all (13 out of 15) leading AI brand websites have their Headline 1 blown up to poster size.

But some have also completely omitted the illustration. Unthinkable in 2016!

How to implement this on your own website:

Promote your headline, demote everything else. Try removing all illustrations from your header.

Make your headline follow one of these proven angles:

• Promise irresistible outcome: "Get better at being healthy, every year" (Superpower)

• Build in a tension: "Build with facts, not vibes" (Dovetail)

• Keep it readable in one breath (5-6 words): “Meet the night shift” (Notion)

2. White background » Hued or dark

Examples: Notion, Granola, Superhuman

A white background can look clean, trustworthy, and Apple-adjacent. For many people, it also looks half-baked and pre-seed.

Another trend past its shelf life is the gradient. Only two of the 15 still pull it off: Lovable and Superhuman.

Half of the 15 analysed brands turned the lights off completely: dark heroes from Superpower, Dovetail, Notion, Cursor, Superhuman, and Sierra.'

The light half didn’t always keep white either. They swapped it for cream and paper tones.

How to implement this on your own website:

Pick a background that matches your ICP’s mood. Tranquil vs. techy vs. experimental.

• Dark = technical, premium, serious (works for dev tools and pro software)

• Cream and paper = calm, human, editorial (works for anything claiming to simplify work or life)

• Gradient = statement. Only if you commit like Lovable’s pink-blue.

Whatever you pick, ensure high text-background contrast first. Dark heroes win design awards and lose conversion tests.

3. Sans » Serif fonts

Examples: VoiceFlow, Ada

I think the sans » serif font change is the best recent UI hack I’ve seen. It immediately creates a classy and elegant feel.

What’s even better: It’s a new trend that most SaaS brands have yet to catch up with. Change now and be among the avant-garde.

Sans serif fonts (Inter, Helvetica, etc.) are safe and neutral. They’re also invisible and identical to everyone.

The taste-signalling AI brands break rank in two directions.

Direction one: serifs.
VoiceFlow goes full editorial, a headline set like a magazine cover. It is paired with a simple sans font in Headline 2.

Direction two: keep your sans font, but drop the weight.

The thin sans is the safe update that you can test without messing too much with your CVI.

Ada’s headline uses a thin sans font which creates a more premium feel.

How to implement this on your own website:

Switch your Headline 1 and Headline 2 fonts. If you use a WebFlow or WordPress template, you can easily change your system fonts.

Five free and classy serif fonts to test:

• Instrument Serif: the current AI-brand favourite

• Fraunces: a quirky old-style serif font

• Newsreader: designed for digital editorial, Conde Nast and New Yorker style

• DM Serif Display: glossy fashion magazine vibes

• EB Garamond: more classic literary publisher font

Unique paid serif fonts: Tiempos Headline (Klim), Canela (Commercial Type), Reckless (Displaay), Editorial New (Pangram Pangram), Ogg (Sharp Type)

4. Show don’t tell » Give don’t show

Examples: Replit, Lovable

“Show, don’t tell” used to be the golden rule of header design: feature a product screenshot and let it speak for you.

The leading AI brands turned the tables. They no longer show but let you test the product before signup.

The header no longer displays the product. It contains it. You can type right into the prompt box, see a quick result, and be transferred into the signup flow.

Remember what I found in the website header copywriting analysis (previous Marketing Fix newsletter): about 90% of your visitors come to website with the conversion decision already made elsewhere.

So your hero’s job isn’t to persuade, but to remove the last step between the visitor and the product.

How to implement this on your own website:

Productise a key feature of your product that visitors can use without signing up. Put it in the hero.

A prompt box, a calculator, a question with 2-3 answer options.

There’s just one condition: the first interaction must deliver value in seconds. A slow or disappointing demo is worse than a CTA button.

5. Imagery » None or all in

Examples: Sierra, Cofounder, Mistral

I’d love to make the Patamon-Angelmon comparison here. But it wouldn’t be quite valid.

Angemon certainly exemplifies the all-bells-and-whistles version of a header illustration. But there’s no room for a Patamon: the modest middle ground is gone. It’s either Angemon or empty background.

Hero imagery in 2026 comes in exactly two flavours: none at all, or art-directed by skilled designers.

Zero of the 15 analysed brands use stock photos. Not a single corporate princess smiling into the camera.

Sierra runs a muted film-still of a real person on the phone. It looks like an Antonioni frame, not a landing page.

Cofounder commissioned a pixel-art landscape, a laptop resting in a storybook forest. Their entire branding is very difficult and expensive to copy.

Mistral went the opposite way: no imagery at all, just massive type and a brutalist pixel pattern.

How to implement this on your own website:

Run the swap test on your hero image: if it could appear on a competitor’s site without anyone noticing, change it.

Two directions to explore:

• All in: real photography with actual art direction, or illustration with an artist’s fingerprint. Commission it instead of licencing.

• None: bold typography on a bare background. Free, and yet stronger than a generic illustration.

6. Social proof » The sacred logo bar

Examples: Replit, Notion, Clay, Intercom

A lot above the fold has rotated over the past decade. Layouts, colours, fonts, CTAs…

One element didn’t move: the logo bar.

However, it is worth noting that the logo bar is a quintessential B2B move. A B2B buyer needs to justify the purchase to their boss. A consumer only needs to justify it to themselves.

In B2C, brands showcase the number of active users or trust site ratings.

Our of the analysed brands, Clay and Intercom do it the best.

Clay stacks the densest bar of the set under a “trusted by 350,000+ GTM teams” line. And note the little “Case Study” banners next to each!

Intercom takes the logo bar to its extreme. Or shall we say “the logo block?”

While not in the header, there’s an entire section of nothing but customer logos right under it on Intercom’s website. Twenty-four of them, Amazon to Anthropic to A24, arranged like a wall of honour.

How to implement this on your own website:

Selling to B2B teams? Keep the logo bar inside or directly under the hero.

• Use a mix of universally recognisable logos + trendy industry darlings

• Six no-name logos are worse than none. If your logos aren’t recognisable, skip them.

• Once your customer list gets long enough, do what Intercom does: give it its own section.

Bonus:

The coolest website I saw when researching this article.

🪡 Needles in a haystack

In this final section, I share my latest cultural discoveries: books, films, exhibitions, seasonal finds. Call it bonus education in taste in the AI era, hehe.

🪡 Books: Recently, a startup founder told me I’d inspired him to read more fiction. One of the books he’d enjoyed was Tomorrow, and Tomorrow, and Tomorrow by Gabrielle Zevin. I’d had it on my radar for years, so I finally picked it up. Reading it feels like watching a good TV series: it pulls you right in. The novel follows two friends as they become world-famous video game designers. The best thing? It reminds you to strive for excellence in everything you do. “If you do it badly, why do it at all?” a character asks. Well, why indeed?

🪡 Poetry: This week, I’ve got two personal literary news. First, my pamphlet of prose poems Skittles Animals has been published by If a Leaf Falls Press. Read more here. And three of my poems appeared in the June edition of Estonian literary magazine Looming.

🪡 Travels: Just a reminder for anyone travelling this summer. Send postcards! It’s a beautiful human gesture that says: “I thought of you and made a real effort to send you this note.”

Grazie mille for being a reader.

In the next issue, I will share some of the best-performing Facebook ad creatives I’ve designed in 2026, and explain how I do video ads as a marketer with no motion design education.

Until then: Please share the newsletter with others who’d benefit.

Share

Need help with marketing?

BIG NEWS! I am open for 2-3 new projects starting in September 2026. Read about all collaboration formats here.

Have a good summer y’all!

Karola

Back in 2016, we called it CRO. Now, we can call it conversionmaxxxing.

Last night, I got a stupid idea for a YouTube series.

Each episode features a 30y+ old startup marketer. We watch the host hand them a laptop. The marketer’s face contorts, either in awe or disgust. What they’ve just been shown is their own handiwork from ten years ago.

How would you feel when seeing the websites, blog articles, and infographics that you created in 2016? Back when cornflowers were bluer and organic traffic 100x higher…

I’m a sucker for nostalgia, so I couldn’t resist testing my own reaction to how Taxify’s website looked in July 2017. That’s when I joined the company now known as Bolt.

Hello, Wayback Machine my old friend.

My reaction was an elated “Awwwwww!” The good ol’ times.

That diagonally split background photo was my arch nemesis for years. It proved surprisingly resilient to any replacement. It just worked so well. The image shows both the app and how it materialises in the physical world. It even perfectly halves the “Tap a Button, Get a Ride” headline.

The next section, right after the hero, explains the key benefit and how the product works. The illustrations are product UI.

AI may have turned many areas of marketing on its head, but the basic rules of the website hero remain intact:

1. Communicate your product’s key benefit

2. Explain how your product works

3. Visually show your product

4. Keep it clear and simple

5. Have one call-to-action

I was surprised to see the current Bolt home page. I don’t have the data, but my marketer nose tells me that for the people looking to order a taxi, their Formula 1 message is confusing af.

It can be argued that Bolt is now a household-name brand, and everyone knows what it’s for. But even so, I don’t get the Formula 1 message. What’s happened to the app and rides?

If someone from Bolt’s marketing team reads this, please reply and explain.

In any case, let us race on towards the leading AI tools’ website hero analysis + tactics for fixing your own.

Your website hero: 2.016 vs. 2.026

The basics of a good website hero section in 2026 are more or less the same as in 2016. What’s changed is how people interact with it.

Thanks to AI and the innovations of BigTech companies, the level of noise in our everyday has increased 100x. Our environments are engineered for interruption, and many of us have the habit of constantly switching.

What it means for your website:

1. The conversion decision has already happened elsewhere.

About 90% of your (human) website visitors come there to convert. They’ve already made up their mind elsewhere: AI recommendations, Reddit, paid ads, friend’s recommendation, etc.

2. Any distraction or confusion might hurt the conversion decision.

If your Home page hero doesn’t match people’s expectations, they get confused and leave.

The trouble with tech products in 2026 is that they do much more than just one key task.

Think of Claude or Clay. They’re great for a million use cases. So how to write a hero headline that conveys it all? How to say as much as possible in as few words as possible?

Both have their own tagline approach. I’ll show you how more AI brands approach the hero headline in a moment.

3. Aesthetics matter more than ever

You don’t have to keep up with every design trend, e.g. 3D icons. But since Apple popularised the Liquid Glass style, nearly all brands jumped on board.

The bad news:

If your website header doesn’t leave a high-quality, trustworthy impression – or looks like Y2K without the intention of looking like early 2000s – visitors move on to your competitors whose website looks better.

The good news:

There are so many amazing AI website platforms and design tools (Wix, Framer, Figma) that enable marketers to set up beautiful, sleek landing pages.

And sometimes, simplicity is the key.

In case you suspect that your own website hero is also leaving the visitors confused – or simply doesn’t convert them – keep reading on.

Website hero tactics of 15 category-defining AI products

One of my convictions that hasn’t changed in time is this: If you want to learn about any area of marketing, see how the best-in-game do it.

And in 2026, the brands with best marketing and design talent are the AI tools.

I chose 15 leading AI products, and conducted a qualitative + quantitative analysis of their website home pages.

My curation process considered:

1. Status as category-leading AI product

2. Reputation for strong marketing

3. Extreme focus on design

4. Angle and category variety

5. Mix of audiences, B2C vs. B2B

The final list I came up with: Superpower, Dovetail, Ada, Clay, Granola, Attio, Cursor, Notion, Lovable, Superhuman, Replit, Wispr Flow, Sierra, Cofounder, and Mistral AI.

As the hero section does 90% of the heavy lifting, I focused on exactly that.

There’s a lot to unpack even in the hero alone:

1. The 1-sentence headline

2. Sub-headline and minimised text

3. Design: light or dark

4. Illustrations: photos or videos or product UI

5. Additional elements: chat interface or CTA, client logos, etc.

The questions I asked myself:

1. How do the headlines convey what the product does, and the USP to the user?

2. What’s the sentence structure of the headline? And the average length?

3. What are the most common elements used in headers?

4. What are the prevalent design trends?

5. What’s in the home page section following the header?

In this week’s Marketing Fix, I’m going to focus on the copy side of things. The next issue will cover design and user experience.

Below are the key takeaways. Read through and apply them right away.

1. Keep your headline short

The leading AI brands keep their website headlines short and sweet.

Ten out of fifteen analysed websites keep the headline at 6 words or fewer. The median Headline 1 length is five words.

The shortest headline award goes to Lovable (“Build something Lovable”), and the longest to Clay (“Go to market with unique data — and the ability to act on it”).

The sweet spot is at the 4-6 word count. “Build with facts, not vibes.” “Meet the night shift.” “Don’t type, just speak.”

But there’s a caveat. The short n’ catchy messages do not work alone.

Simply reading a headline like “Build with facts, not vibes” or “Meet the night shift” wouldn’t tell you what the product does.

The short headlines always have a subheadline underneath them.

This part carries the literal description. As many as 13 of the 15 sites split the copy exactly this way.

Still, I’d argue that the best website headers skip the subheadline completely. More on how they do it shortly (pun intended).

2. Most headlines talk about brand and/or category

It doesn’t matter whether your headline is five or twelve words, if you fail to convey what your product does + how it benefits the customer.

Of the 15 sites I analysed, the Headline 1 fell under one of six angles.

1. Brand as a subject (4 of 15): X helps you with Y

• “Cursor is your coding agent for building ambitious software.” (Cursor)

• “Cofounder lets you run an entire company with agents” (Cofounder)

• “Better customer experiences. Built on Sierra.” (Sierra)

• “Build something Lovable” (Lovable)

2. Category claim (3 of 15): What is this product?

• “The agentic customer experience platform” (Ada)

• “The revenue platform engineered for scale” (Attio)

• “The AI notepad for people in back-to-back meetings” (Granola)

3. New category defining (3 of 15): Meet this new product

• “Superpowers, everywhere you work” (Superhuman)

• “Frontier AI. In your hands.” (Mistral AI)

• “Meet the night shift.” (Notion)

4. Contrast (2 of 15): Do X not Y

• “Build with facts, not vibes” (Dovetail)

• “Don’t type, just speak” (Wispr Flow)

5. Outcome promise (2 of 15): Do or get X

• “Get better at being healthy, every year” (Superpower)

• “Go to market with unique data — and the ability to act on it” (Clay)

6. Question (1 of 15): Question + chat interface

• “What will you build?” (Replit)

3. Put the blablabla in subheader

Nobody reads it anyway.

And yet, most analysed websites still do it.

My bet is that it’s not for people, but AI algos scanning the websites to understand what the product does.

As to humans, the better place to address them is in the section right after the hero.

Several analysed AI brands use the after-header section to communicate how their product works.

4. Test non-traditional CTAs

The good old “Get Started” still works: 8 out of 15 brands run some version of the classic formula.

• “Get started” (Cursor)

• “Start for free” (Attio)

• “Try Dovetail free”

• “Download for free” (Granola, Wispr Flow)

• “Get Notion free.” (Notion, duh!)

“Try” and “Start” are the default for a reason: it requires zero thinking from the visitor. No objections = more conversions.

The more creative (but not necessarily better-converting) heroes treat the button as one more line of copy.

If your headline angle is strong, the button is free real estate to say it again .

• “Run a company” (Cofounder

• “Become a member” (Superpower)

• “Send me a demo” (Attio)

• “Listen to a real call” (Ada)

• No button, just a prompt box (Lovable, Replit).

In 2026, the most radical version is the CTA that is the product. Replit pairs it with “What will you build?” and provides a textbox field.

Personally, I’m a fan of interactive headers that invite the visitor to engage with the product before taking any action. It pulls them right in.

5. “AI-powered” is SO passé

In 2023, having AI in your product was a novelty. In 2026, it’s self-evident.

And yet, every marketer still faces the Headline 1 dilemma: to say the magic word, or to assume that everyone knows, anyway?

As to the leading AI brands, one of the last words they’d use in their Headline 1 is “AI.”

Only 5 of 15 analysed headlines mention AI or agents at all. But 12 out of 15 still say it somewhere above the fold.

Sierra, Replit, and Superpower never say it anywhere above the fold; Superhuman tucks it into the sub.

1. “AI-powered” is dead.

Zero out of 15 headlines use the 2023 cliché. The word either moved up (”Frontier AI. In your hands.”), or moved out entirely.

If AI-powered” is still in your headline, get rid of it fast.

2. “Agent” is the new “AI.”

Cursor sells a “coding agent,” Cofounder runs your company “with agents,” and Ada claims an “agentic” platform. Notion and Clay tuck “agents” into the subheadline.

The framing shift is significant: AI describes a technology, agent describes a hire. Further, “your coding agent” instead of “an agent” turns a software tool into staff.

3. Omission is a status symbol.

Sierra, Superhuman, Replit, and Superpower skip AI in the headline entirely. Sierra doesn’t even say it anywhere above the fold. When you’re the category leader, AI is assumed.

Increasingly, putting “AI” in your headline reads like a redundant buzzword.

TL;DR: Five key takeaways

Research takeaway #1: 10 of 15 leading AI brands keep their headline at 6 words or fewer.

Research takeaway #2: Most headlines follow a brand or category led angle.

Research takeaway #3: The subheadline's real reader is increasingly a machine. Humans get convinced in the section after the hero.

Research takeaway #4: 8 of 15 brands still run the classic "Get started / Try free" formula. The boldest replace it with a prompt box.

Research takeaway #5: Only 5 of 15 headlines say "AI" or "agent" at all, zero say "AI-powered," and the category leaders skip the word entirely.

Test your own website hero against these five. If it survives, reply and let me know.

In the next issue, I’ll be covering the design part of the analysis.

Light vs. dark heroes, product UI vs. illustration, and why nobody uses stock photos anymore.

Needles in a haystack

I’m sorry, but I’ll have to skip the needles this week. Not least because I’ve been so busy w/ work that there’s not much I’ve read or seen. Oh well… [Questions her life decisions]

Grazie mille for being a reader.

I’ll be back in two weeks with another Marketing Fix.

Until then: Please share the newsletter with others who’d benefit.

Share

Need help fixing your marketing game? Read more and contact me here.

Have a good summer y’all!

Karola

If you know someone who’d benefit from the free Audit Questions Library, please share this post with them. They’ll get the link to Notion file in Welcome email.

Share

Before getting into the A-Z of it, I’d like to note that…

Summer is one of the best times for conducting a marketing audit.

Half the team’s off, and so are your customers.

Slow office days mean lower distraction. And this is just what’s required for a good, productive writing day.

Worried about not having enough experience?

That’s just another excuse. As with many things in life, Google is your best friend whenever in doubt.

Give it a go!

What is a marketing audit?

After working on 45+ marketing audits, each of which takes 20-40 hours, I’ve spent ca 1,350 hours of my waking life (that’s about 84 days!) switching between tabs and docs.

How did I end up writing the audits? And why?

It’s rare for agencies or consultants to offer marketing audits as a standalone service. They’re usually done in-house, by the CMO. Or by an external auditing agency as part of a bigger project.

And yet, the need for an audit of a company’s marketing activities arose every time I started working for a new brand.

A 360-degree marketing audit is the fastest way to:

1. Get an exhaustive overview of what’s worked / not worked.

2. Understand each marketing channel’s state and ROI.

3. Research competing brands’ USPs and tactics.

4. Map new ideas and opportunities for growth.

5. Align company-wide goals with those of marketing.

6. Get a prioritised roadmap for the next 6 months.

Think of a marketing audit as a complete check-up of your marketing strategy.

Audits are done 1-2 times per year, and give you a deep dive into your marketing efforts to see what’s working and what’s not, and what could be optimised for better results.

You end up with a clear roadmap for the next 6-12 months.

What does a marketing audit cover?

When conducting marketing audits for my clients, I usually spend 20+ hours collecting, analysing, and formatting the information.

The resulting Google Docs file is about 60 pages long (including screenshots).

That’s a lot of hours and pages…

But if you think about it, there are also many different aspects to marketing. A well-researched and actionable marketing audit covers every aspect of your marketing strategy, from acquisition to engagement to retention.

The best way to do a marketing audit is by taking the 360-degree approach.

You should cover all the key aspects of your growth and marketing strategy, from planning and reporting to detailed per-channel analysis.

Eight key chapters of a marketing audit:

1. Results & Reporting

2. Brand & Value Proposition

3. Website/App & Acquisition Flow

4. Customer Acquisition

   1. Growth loops

   2. Paid ads

5. Social Media Marketing

6. Content Marketing & SEO

7. Email Marketing & Automation

8. Summary & Next 12 Months’ Plan

How to get started?

1. Define your marketing audit goals.

2. Set up a Google Doc with all chapters.

3. Block 2-4h every day to complete the audit in 1-2 weeks.

1️⃣ Start by defining what you want to get out of your audit.

• Is your goal to improve ROI or find new ways to scale?

• Are you looking to identify weak points in your strategy?

• Just checking if your marketing strategy aligns with the CEO’s view?

Be clear about what the final audit result should look like before you start.

2️⃣ Set up a Google Doc with all chapters.

The hardest part is to start the project. Just do it.

I like writing marketing audits in Google Docs, as they’re great for formatting, including screenshots, and team comments.

Alternatively, you could use Notion or a Miro board.

3️⃣ Block 2-4 hours of time daily.

Writing a marketing audit is not a multitasking side task. Your brain needs to go into deep-focus mode where it can take in lots of information, analyse it, make connections, and come up with strategic takeaways.

I usually aim to write a full chapter in one sitting. It ends up saving a lot of time, because you don’t have to reopen all the tabs, get back into the mindset, etc.

It’s best to work on your marketing audit in the morning, when your mind’s not yet cluttered with Slack threads and meetings.

Can AI save you time?

There is one part of the marketing audit where AI tools like Claude Cowork can save you time.

That’s competitive research.

I like to create a list of 10-15 competitors, and ask Claude to scrape their website and socials, then create a spreadsheet with relevant URLs, positioning and value props.

But even then, I’ll manually click each of the URLs and analyse their websites and social posts to know what’s actually there. I also collect screenshots of good examples (website blocks, USP wordings, paid ad creatives) and add them to Google Docs with my own commentary and takeaways.

As to the rest of your audit, you’ll only waste time trying to cheat with AI.

The whole point of the marketing audit is to view, analyse, and process everything with your own wise human brain.

And, obviously, you’re the only person who can come up with meaningful takeaways and next steps.

Use AI cautiously. Otherwise, you’ll end up with lots of gibberish that does not lead to any improvements or change.

On to the Notion Library!

Access the Notion library here: Free Library: 95 Marketing Audit Questions

Copy it to your account for tracking your progress. Tap “Duplicate” icon in the top-right corner.

Needles in a haystack

In this final section, I share my latest cultural discoveries: books, films, exhibitions, seasonal finds. Call it bonus education in taste in the AI era, hehe.

🪡 Books: At a recent Soho Readings event, hosted by my dear friend Tom, I picked up Permanence by the contemporary British writer Sophie Mackintosh. It turned out to be a fun, easy read. The narrative oscillates between the real and Other world, while a couple in love figures out whether they can ever really be together.

🪡 TV & film: I finally watched Nouvelle Vague, a 2025 homage to the French New Wave film director Jean-Luc Godard. What I liked most were the constant references to genius, and how one must strive to lure it out. Like a rare glimmer of sun in a cloudy sky. Pretentiously, a character quotes Sartre: ‘Genius is not a gift, but the way a person invents in desperate circumstances.’ It’s refreshing to think of peak human achievement, when everywhere around you slop and sameness abound.

🪡 Arts & research: There’s an exhibition of one my favourite painters, Francisco de Zurbarán, at the National Gallery in London. For all the mystery and drama of his saints, it’s his quiet and sumptuous still lives that most dazzle me. Did you know why the 16th-century Dutch still lives featured opulence (gold and oysters), while the Spanish depicted simple fruit and vegetables? That’s because the latter country was severely Catholic, and any public show-off of one’s fortune was considered vulgar.

Grazie mille for being a reader.

I’ll be back in two weeks with another Marketing Fix.

Until then: Please share the newsletter with others who’d benefit.

Share

Need help with marketing? Read more and contact me here.

Have a good summer y’all!

Karola

If you’re not yet subscribed, you can do it here. Or say hi on LinkedIn.

Subscribe now

I put off starting this article on marketing AI armageddon for over a week. The mental overload from infinite AI use cases is real. To the point that, finally, to give this sense of chaos a tangible image, my brain time-travelled 2 million years back to Pleistocene.

The image that arose in my head at 9pm on a Wednesday evening was of a woolly mammoth.

A dead woolly mammoth, to be exact.

(Certainly not what ChatGPT would suggest. All for the better, perhaps…)

So here we are, in the early Pleistocene.

The planet is much warmer and wet. All across African savannahs, herds of proboscideans (mammoth ancestors the size of small trucks) are crunching on shrubs and twigs.

Then, sadly, one of them is taken down by an ancient pathogen.

There it is, laying in the tall grass, waiting to be eaten by the first Homo habilis to come upon a dead mammoth.

Imagine a hairy human ancestor—let’s call him Charlie.

Out on his daily hunt, Charlie finds the brown mountain of meat. He doesn’t know it’s a miniature mammoth. What he does know is that his fifteen children are hungry, and animal meat very yummy.

Now, put yourself in the shoes (pardon, an anachronistic pun) of Charlie.

You’re a tiny human standing beside a huge mammoth cadaver, unsure what to do about it.

At first, you walk a few circles around the animal, contemplating not whether, but how, to break it apart. You shrug, then shiver at the thought of having to tear the whole thing to pieces and carry it home. It would take you several days, 100 kilometres back-and-forth. You’d rather leave it there, go home to a cave and bring it up again.

Soon, it will all come to seem like a distant bad dream.

The AI marketing mammoth

The mammoth is a fitting metaphor to describe how most marketers feel about AI.

It’s something supposedly good. But it’s also horribly big. Too big to make heads or tails of it.

And yet, you read another LinkedIn post and decide to check out Claude Code. First, you shrug. Then, you shiver. And because it’s a capitalist tech-driven world, you take out your stone dagger (trackpad) and cut (tap) into the pelt (signup button).

Yuck.

This was my first thought, too, when downloading Claude Desktop 6 weeks ago.

Then, I took some days to read my favourite marketing newsletters’ recent archives. When I returned to the real world, my eyes were sore and my head haloed with an artificial 3D ring.

The AI overwhelm is real

The key reason for marketers being stressed out over AI is that we look at it as one huge beast.

While, in fact, it’s an entire marketing team.

Just like a mammoth has legs, trunk, and belly, your marketing team is made up of several specialists.

But read five posts on your LinkedIn feed, and your mind reels. AI-led sales outreach, blog post generation, ASO optimisation, marketing asset design, etc. etc. etc. And you, feeling like I need to do it all NOW.

Because only a loser couldn’t implement 10 commandments by 10 diverse marketing experts in one week. Because only the bad marketers, soon to be replaced by AI, would fail to set up 100 new Claude Code skills and foundations and grow their company’s blog traffic + inbound leads 700% MoM in three months.

Right?

Ummm… There used to be LinkedIn posts about CMOs explaining to CEOs why this can’t be done.

I want to return to that pre-AI world.

But of course, I can’t.

To stay competitive as a marketer in 2026, you have to adopt AI to some extent.

Deconstructing the AI marketing mammoth

The most helpful thing for me, when returning to the marketing realm after 8 months of writing my sci-fi novel, was to deconstruct the AI machine.

It’s really not that different from joining a new company as marketing lead. Which is exactly what happened as I joined Roleo as their part-time founding CMO.

If you’re a full-stack marketing lead, you want to do (read: rescue) everything at once. There are obvious fires to put out. Meanwhile, you’re supposed to build a long-term marketing strategy, set up KPI tracking, launch paid advertising, and set up social media accounts. Not to speak of product launch support and website revamp.

It’s the same way with AI.

The abundance of AI use cases in marketing leads to decision fatigue and paralysis.

According to thousands of LinkedIn posts by the “experts” who use ChatGPT to scrape and rephrase their “expertise” into “comment for skills” posts, you should use AI for everything.

• Competitive research

• Daily & monthly planning

• Paid ads design & optimisation

• Content marketing & ASO

• Website & assets design

• Outbound sales outreach

• Social media growth

• Media outreach & PR stunts

The advanced gurus even suggest that you to build your AI second brain.

Sorry, Kieran. While I consider you one of the best marketing AI experts, I choose to keep my single brain and plan my own life.

The best way to overcome the sense of helplessness is to break the “AI in marketing” project into small, digestible pieces.

Marketers who stress about AI are looking at it as a single project. The solution is to break it apart and tackle it project by project.

Instead of saying: “I need an AI control centre for marketing.”

You could say: “I’d like to set up Claude Cowork to help me with competitor research and blog article writing.”

What changes is the project’s magnitude. By verbalising the exact outcome, you give yourself the agency to finish it.

1. Identify 2-3 marketing AI use cases that would benefit you most (time savings or more quality output).

2. Create a timeline for building out a Claude Cowork (foundation + skills) project for each.

3. Take 3-4 weeks to build the Claude project, test it, teach it, and start benefitting from it.

Let the final point settle in. Setting up Claude Cowork for high-quality output takes time. It can’t be copy-pasted from a LinkedIn post comment.

Which AI marketing use cases should you explore first?

It depends on your role in the company, and how advanced your marketing game already is.

Further, there are still some use cases where AI simply can’t replace human intelligence, creativity, and skills. Like designing ad creatives or writing opinion essays. I didn’t use any AI to write this newsletter, only for spell-check later.

The image below shows how GTM teams use Claude Code. (GTM = product marketing, sales, growth, lead gen.)

Take 30 seconds to consider where AI is most useful for you. (Remember: you are one person, not a team of twenty.)

As a full-stack marketer working with early-stage companies, my needs are quite different from the above chart. I don’t work with sales; I have to deliver immediate growth.

If you’re a mid-level marketer in a large team, specialised in 1-2 things, let me tell you this. Big companies hire automation experts to set up complex AI systems, including for marketing tasks. These people are paid €250k/year plus.

If you’re in a team of 5+ people, it should be your CMO, not you, who sets up Claude Cowork for the marketing team. And ensures it’s aligned with the rest of the company.

How I’ve been using Claude this far

Here are the primary tasks I've used Claude for over the past six weeks, and where I found it helpful vs. adding to the chaos.

1. Competitive research: 8/10

Claude saved me 80% time by collecting all information in a spreadsheet. I still had to specify the competitors to check, collect relevant screenshots, review, analyse, and summarise.

2. Product marketing copy: 6/10

This use case only became useful after I set up a Claude Cowork project with detailed foundation files for brand voice, product features, competitors, and audience. It took me 3 working days and was totally worth it. I use it for drafting website and product launch copy, then rewrite everything in the end. It’s great for creating a logical structure and covering all important points. Also, for brainstorming 50 different headlines with a slightly different angle and wording.

3. Social media ideas & copy: 5/10

Horrible for original ideas. But once you set it up as Claude Cowork project, it can help with the first draft. Create brand voice guidelines, and provide detailed, original input. (No dropping in 50 LinkedIn posts by other people and asking the algos to copy them. In gastronomy, you can turn stale bread into delicious pudding; it doesn’t work in marketing.) In the end, I always rewrite the whole draft and use Claude for review.

4. Meta ads copy: 7/10

Actually pretty good, once you’ve set up Claude Cowork (see point 2).

5. Meta ads design: 1/10

A complete waste of time. Don’t believe anyone who tells you otherwise. Even the AI-design oriented tools like Figma, Adobe, Claude Design can’t properly render logos and align text. No shortcuts: learn Figma or hire someone.

6. Ideation & research: 5/10

I tried asking Claude for fun Latitude59 tech conference marketing stunts. It suggested some ok stuff. But ok is not good enough. In the end, it took me a 2h walk around East London to come up with the actual idea.

7. Charts & graphs design 8/10

I’m giving it a high rating because creating the visual below took me 10 minutes with Claude Design, from project setup to download. Caveat: I had to prepare a Figma file with brand assets and chart examples, which took me 2 hours. Plus you’ll need a CVI. Even so, huge time savings ahead.

While many pros recommend using Claude Code, you can do most marketing work with Claude Cowork. It’s got chat-based interface instead of Terminal.

“Just three months after launch, Cowork (32%) has already surpassed Claude Code (31%) as the primary Claude product among GTM operators we surveyed,” Maja Voje and Kyle Poyar reported in April 2026.

The survey of 200 GTM specialists confirmed that many prefer Cowork, especially for content creation, product and growth marketing.

Where I wouldn’t use AI as a marketer (and a human being)

Roughly 75% of the AI use cases the LinkedIn experts hype, I’d never use myself.

They either take more time than they save, or produce low-quality results.

1. Reporting & ad account optimisation: I prefer to do it manually every week, and process the information with my own brain.

2. Website & ad design: It’s just not working yet. I made some graphics with Claude Design for this newsletter. Not sure I ended up saving time.

Meanwhile, I was working on rebranding Marketing Fix this week, and did everything manually on Figma. Claude was helpful for researching premium typefaces. I then browsed are.na for hours to research colour swatches, editorial style, and typography. Finally, I designed everything manually on Figma.

3. Social media & newsletter writing: I’m a writer, after all. And I find something deeply degrading about serving other human beings “content” written by a machine. Just as I wouldn’t serve my friends Snickers and McDonald’s fries.

4. Human-to-human communication: I understand why sales teams contacting 100s of people each week automate the process. You can do it with Apollo, Clay, etc. But as I only send 5-10 work-related texts per week, I prefer to write them myself.

The stuff you don’t need AT ALL

I’ll end my rant-slash-therapy-session by repeating my unswayable belief:

There are many overcomplicated marketing AI use cases nobody actually needs.

Here’s an example of the type of LinkedIn post we’d all be better off without.

What it does is expand a single Claude Cowork use case into 20 “agents.”

Most of these are useless to your decision-making. They only add noise and extra work.

And extra anxiety.

Instead of creating 20 AI agents, you could simply set up a Claude Cowork project with a solid foundation and 5-6 skills for ideation and content creation. It writes all the assets in one task, instead of being uselessly broken down into 10 different details. Plus you might set up some bot for reaching out to collaborators and answering comments, if the volumes are high.

The branding, reporting, and designs you still have to do manually. That is, if you want them to be any good.

How to get started?

You probably already have.

It’s me who returned to marketing after a hiatus, and had to begin learning from scratch.

But in case you want a reset, or are still in the overwhelm phase, I recommend that you start following 4-5 people who are real experts on the subject.

Here are my go-to Substacks:

• Maja Voje: GTM Strategist

• Kieran Flanagan: Marketing AI in Action

• Ruben Hassid: How to AI

I will also be sharing detailed breakdowns of how I use AI for marketing tasks in this Substack.

Help to keep me motivated by inviting other marketers and founders.

Share Marketing Fix

Coding skills vs. taste

There’s a new discourse about taste being the AI era’s most valuable skill.

I agree.

When anyone can generate marketing text and visuals in a single prompt, curation and authentic ideas become key differentiations.

As a marketer, you should be well aware that as long as your product is alright, original branding is what makes it stand out.

What can't be prompted is something that didn’t exist before you created it.

“Writing, good writing, will therefore always be a place where something unknown, something which didn’t exist before, is given existence,” the Nobel Prize-winning Norwegian author Jon Fosse wrote.

Instead of sweating about learning to code, become genuinely curious about art, dance, books, film—whatever you fancy.

Spend time away from your screen, in the real world. Because the people you’re marketing to, god bless, are still not running on AI brains.

Here’s a list of the best books I read a few years ago. Perhaps you’ll find something to get you started.

Needles in a haystack

In this final section, I share my latest cultural discoveries: books, films, exhibitions, seasonal random finds. Call it bonus education in taste, hehe.

🪡 TV & film: I usually stay away from Netflix, but my friend got me hooked on Season 2 of BEEF while I was visiting him in Berlin. Watch it for suspense, laugh out loud, and insanely great acts by Carey Mulligan and Oscar Isaac.

🪡 Books: One of my fave contemporary writers Ben Lerner had a new book out in April. Transcription is a 130-page novella that explores tech-infusion in our lives, and how our memories can differ from each other. Here’s a very good review of the book.

🪡 Arts & research: An artist friend recommended are.na to me as a better version of Pinterest. And it certainly is. Check it out for brand, fashion, and other aesthetic research. I’ve found amazing boards with magazine typography, 90s Prada editorials, and colour swatches.

Substack’s giving me a warning of this dispatch being “near length limit".”

I’ll be back in two weeks with another Marketing Fix.

Until then:

1. Do let me know what you think of the AI marketing topic, and what you’d like me to cover next. Just hit reply.

2. I’m keeping Marketing Fix free for the next five issues, perhaps longer. So feel safe to share this post with friends who’d benefit from some AI therapy.

Share

Have a good spring y’all!

Karola

Last October, I went quiet for a while to write my debut novel.

The first draft is 3/4 done now, I’m at 75k words (ca 280 pages). I’ve also written half a collection’s worth of short stories & my prose poetry pamphlet will be out in late June 2026.

What I didn’t expect was how much I’d come to miss marketing. So I’m back in the game much sooner than I thought. And so is Marketing Fix.

The newsletter has moved to Substack, which is why this email looks different.

Here’s what to expect from now on, bi-weekly on Fridays:

↳ Hands-on marketing tactics

… that make you better at copywriting, landing page and ad creatives design, marketing audits, paid ad strategy, and more. Plus free templates here and there. E.g. 80 Marketing Audit Questions: Free Checklist

↳ Hot-take reflections & predictions

Opinionated takes on the latest marketing trends. Plus the occasional takedown of stupid things, AI slop chief among them. E.g. I Gave Up on LinkedIn. Should you? + 55+ Predictions for 2025 Marketing Trends

↳ Needles in a haystack

My latest cultural and literary discoveries I can’t resist sharing: books, films, exhibitions, seasonal random finds. Call it bonus obligatory education, hehe. Officially my favourite section of Marketing Fix.

The first new issue lands very soon. If you have a marketing question you’d like me to cover, DM me on LinkedIn.

Have a good spring y’all!

Karola