Overtone will be out and about this autumn. Come and find us at:
PubAcademy in London
IBC in Amsterdam
AI Summit in Barcelona
Prebid DevCon in Cologne

Philip wrote a reflection of what was different at Cannes Lions this summer

Insights to Action

This year’s Cannes Lions was trying to get back some of the glamour in the era of agents, trying to get at something bigger and more creative than whatever the next token says we should do.

YouTubers roamed around the conference in the Palais, the talent agencies like UTA and CAA each had big tents, and the sweaty AdTech hordes did their best to look good while melting in the heat of Southern France.

One place that was trying less hard was the insights track of Cannes Lions, where researchers from major companies like IKEA, or Kraft or Warner Bros gathered in a nondescript conference room to chat over what would become of their profession. Previously they could spend months doing academic-level research into what their audiences felt about a particular subject or product, where they hang out and what sort of music they like. But how do big manilla folders of surveys survive in the age of instant research done by Claude?

Many of those there complained about having to deal with budget cuts while their colleagues out on the beaches splashed cash on new AI tools. However, one senior exec at a major brand said that she could see things turning their way.

In an age of agents, why shouldn’t all of marketing, and all of communications even, be downstream of insights?

Knowing your audience, their user needs (shoutout to Dmitry Shishkin) and what makes them tick has become a common refrain in audience work for publishers. But it is ultimately something that unites all parts of the media ecosystem from bards to bean counters … albeit with one big problem

The Particular: GPT Ads

Huge amounts of data now powers the idea of knowing your audience, from newsletter opens to less savory personal data floating around the ad space. But all of this data is, by definition, from the past. That is part of the way machine learning, as well as LLM responses, works: training on previous data and finding patterns in order to predict the answer.

The media world, both for publishers and for advertisers, moves at the speed of now, and raw models don’t have access to the latest information. You need something that gives you up to date accuracy. You need news.

One of the places where news is particularly useful is in ChatGPT, where users are asking millions of queries daily, including on what is happening in their communities. This fits right into what Overtone does understanding news, so we decided to test our “contextual intelligence” with the hot new thing in town, GPT Ads. These are now live in North America, Asia, South America and starting this past Monday, in most of Europe.

GPT Ads work targeting ads on conversations from users based on what OpenAI calls “context hints”. People have often been approaching these as similar to Google keywords, though the difference between a Google search and a GPT conversation in terms of format is vast, with the LLM conversation spanning potentially thousands and thousands of words.

Overtone conducted an experiment this summer with our own brand and compared running an advertisement for our news skill with “keyword” style targeting vs targeting conversational “moments”. The hypothesis was that our potential audience was reading the news, and that ChatGPT was using the work of publishers (payment is a separate issue) to answer questions about that news.

So we ran two campaigns for ourselves, the first targeting broad keywords like “Artificial intelligence” and “Claude Code.” The second was based on output from Overtone news analysis of moments that would resonate with our crowd of tech-savvy media types, like the people who read this newsletter, and included the moments such as the rapid pulling of Anthropic’s Fable 5, the release of the Reuters Digital News Report and the launch of the AWS Website Application Firewall for publishers. The campaigns were given the same budget and the same rules for “cost per click” of $3, running for a few days.

The results were clearly in favor of moments, where the clickthrough rate was 1.6%, which according to estimates is in the top 10% of all ChatGPT campaigns. The generic campaign, however, could not even complete its targets because $3 was too low to compete with all the other people buying the keyword “AI,” and we had to double it to get any results at all. That campaign ended up at a click through rate only two-thirds of the performance of the moments campaign.

We are now doing more tests across different campaigns, but for now this is an effective way to meet your audience where they are, in the particular moments where they are interested in a subject you care about. All that is thanks to news, crafted by journalists who care about accuracy and their audience. News also updates every minute.

One example that came to mind was the World Cup. During June and July, thousands and potentially millions of people asked their handy LLM to explain Norwegian rowing chants or who goalkeeper Vozinha was. It is now possible to interact with people in those moments, across articles, videos, and now through their AI assistants.

The Universal: Ulysses and Don Draper

The future of GPT ads, the platform’s relationship with publishers and the ads’ usefulness to normal human beings is still in flux, like the information environment itself that is constantly changing.

I think that trying to understand the big unknowable world as it twists and turns is going to be an increasing part of what every business does in an agentic era, and the advertising/marketing role acts in some ways as a navigator on the ocean of content. It can be scary, a bit reminiscent of the Odyssey, hopping from island to island without knowing if you are going to be attacked or bewitched.

I may not have the writing talent to really connect the Odyssey to modern day media. But luckily I don’t have to, because Irish literary giant James Joyce already did in his masterpiece Ulysses.

The main character, Leopold Bloom, who works in advertising, walks through early 20th century Dublin, bombarded with images and sounds not unlike being on the internet in the year 2026. He tries to make sense of life, death, sex and kidney pies, with Joyce using more than 500 stream of consciousness pages to tackle big existential questions about meaning.

The world traveler Joyce, when asked why he always wrote about Ireland, once said, “I always write about Dublin, because if I can get to the heart of Dublin I can get to the heart of all the cities of the world. In the particular is contained the universal.”

All of this, navigating through layers of meaning, can seem quite heady until you remember that it has always been an essential part of media.

Take everyone’s favorite 20th century ad man, Don Draper of Mad Men. He is often called to mind as part of a bygone era of big budgets, creative panache and untold numbers of cigarettes.

But the show is the exact opposite of a Norman Rockwell illustration of life in the 1960s, it is about the tumult of America during that time, a cascade of moments from the Beatles to assassinations, and how characters all move through history in their own way.

This takes place through the personal, professional and political, all culminating in Draper thinking up the famous “I’d Like to Buy the World a Coke” hilltop advert. He takes everything that has happened over a decade and connects the particular to the universal.

Navigating the ocean of content, as history advances one minute at a time, is also what we are all called to do both in publishing and in advertising. While new tools can create a flood, we can also use our knowledge, our insight and our humanity to reach and connect to people in honest authentic ways, helping them see things in a new light.

Imagine a computing event without Apple. A sports exhibition without Nike. A banking conference without JP Morgan.

Allow me to describe Cannes Lion 2026 to you in the buzzwords I counted: AI, Agentic, rosé, creator, AI, yacht, vibe-coding, sun, consolidation, AI, party, creativity, heat. AI.

Which is why I had a growing, uneasy sense during the week that something obvious was missing. None of the titans of AI were really represented as such. Yes, Google was there (with a cute little ‘Gemini’ banner) but they are part of the furniture. Yes, OpenAI did a little speech in the Palais about its in-house ad tool. But that was it. No beach space for Anthropic, Perplexity, or, heaven help us, Grok and Palantir.

Here’s my unlikely-to-be-popular opinion. If an AI company had custom-engineered a festival to persuade the ad-tech industry into self-sabotage, they couldn’t have done a better job than Cannes 2026 did to itself.

There were hundreds of sessions by publishers, agencies, holdco’s, and tech providers gushing over technology.

“Productivity is skyrocketing!”

“Of course we’re cutting staff: everyone is.”

“Have you heard about [insert acronym]? I don’t understand it but it’s amazing.”

But: they completely ignored the industry’s exposure to existential workflow risks.

There’s no such thing as a free lunch. Except on the Croisette, where incumbents happily feed the humans who are feeding the machines who will feed on those incumbents’ profits. The ad-tech industry is doing an excellent, and largely free, marketing and R&D job for the AI giants.

Agencies and holding companies acted as pro-bono evangelists for companies into which they shovel internal and client data. These are companies who spend double-digit percentages of their revenue on enterprise marketing, but now it seems they are voluntarily spending a week on the Croisette pitching each other on the importance of integrating generative AI tools, immediately.

There’s been a breathless rush to look innovative (and show the bosses that your balance sheet is competitive). In that whirlwind, it’s easy to overlook legal complexities, but perhaps the more important risk isn’t when rules are broken, but what happens when the rules work exactly as intended. We are teaching the machines how to do our jobs, one marketing campaign at a time.

To be fair, this wasn’t the whole festival: and it’s worth being precise, because ‘advertising’ and ’ad tech’ are different beasts. The creative side of the industry showed up in spades to defend itself. There was a Palais track built on the premise that creativity should have an increased role, with Chipotle’s Fernando Machado warning against ‘confusing activity with impact’. The people who make the ideas have already asked whether they’re automating away their own relevance. And they answered it, by insisting on the distinction between the pipes and the creativity, and betting that you can’t machine your way to the latter.

It’s those pipes I worry about. This is the domain of ad-tech: flows of data and the media-buying machinery. But no one was asking that question at all. There, the mood was enthusiastic. A People inc. exec, quoted by Alessandro De Zanche, called it ‘the first post-ad tech, post-open web Cannes’.

The handful of panels I caught celebrated ‘efficiency’, ‘scale’, ‘optimization’ and ‘profit’. Absent was any discussion of the structural erosion of the ad-tech model itself. In using these third-party tools to scale operations, the plumbing side of the industry is providing the training data, in the form of human strategy, cultural context, even creative insights, to tech platforms that may eventually disintermediate the agencies. And the holdco’s. And the existing tech platforms.

A stroll through Cannes in straw hats and pastel shorts, chatting to industry contacts and the occasional superstar, and the, admittedly unsexy, question of ad-tech plumbing is an easy one to miss. The creatives were loud, the pipes didn’t pipe up.

But the industry is ignoring historical precedent. Ten or fifteen years ago, agencies handed their data and distribution keys to Google and Facebook, sidelining themselves into lower-margin knob-twiddlers. Now, they risk doing the same thing with the LLM giants. To add insult to injury, this time they aren’t just outsourcing the distribution but the thinking.

Engineers roll their eyes at sales reps vibe-coding products together. Now the ad industry is vibe-coding its long-term corporate strategy, hoping everything just works out because the energy feels right.

Every panel, and plenty of conversations, were peppered with the word agentic, commonly understood to mean ‘I outsource as much as I can to an LLM’, but commonly implied to mean ‘I, a real human, now have more agency to do great work I didn’t previously have time for’.

This brings us to the fundamental architectural shift of the media landscape, echoing Christopher’s framing of Forts and Ports.

The Old World was an empire of custom Forts: hard-coded integrations between legacy CRMs and proprietary APIs, requiring armies of expensive engineers to defend, maintain, and navigate the manual plumbing. That complexity was the agency’s moat.

The Agentic World, by contrast, operates on Ports. All an AI agent needs is an objective, permission, and a place to dock. It navigates user interfaces and interprets data on the fly.

As a newcomer to the ad-tech industry, I saw as its saving grace the complex human/tech blend: messy plumbing of APIs and enterprise software that required people power to orchestrate and sell. But in an agentic world, software integrates itself. The agent is the integration.

So it’s worth being precise about what actually gets replaced. This isn’t agents ‘replacing agencies’ wholesale. Creative shops that sell great ideas will stick around. But the integration and orchestration layer that thins out first, because it’s so easy to automate and scale.

The holdco’s are no longer really (groups of) agencies in the traditional sense. They’ve spent a decade turning themselves into technology companies. When the integration work evaporates, their position as the ‘stewards of complex systems’ disappears with it.

Rather than jump between ad-tech and agencies and creatives, it’s cleaner to draw a line. On one side, the companies that will respond by trying to become technology companies themselves. Besides ad-tech firms, that includes holdcos: WPP’s Open Intelligence, Dentsu’s AI Playground, and so on. On the other side sits everyone working in a more open, agent-native system, where you don’t have to own the integration to get the work done. The question is what leverage the first group still has. Established companies were built on billable hours and headcount, and for all the reinvention, they still own neither the data nor the infrastructure (although Publicis seems to be making a play for this with their LiveRamp acquisition). Once brands realise they can reach the intelligence and run campaigns directly, what keeps them paying a toll to the middle?

The winners, for now at least, are surely the AI behemoths. Their presence at Cannes was ghostly, but their impact is, and will continue to be, very real. Meanwhile the vibe was (as, I suspect, it always has been) strong. Enthusiasm is a great creative tool, but it is a god-awful business strategy. It’s time to look past the short-term efficiency gains, before the vibe decimates the business model entirely.

Food for thought. Human thought, please.

Perugia, Italy, and Cannes, France are 400 kilometers apart, and worlds away from each other.

They actually have many similarities. Both have been around for thousands of years, evolving from villages to fortresses to tourist destinations. Both have changed hands between warring parties countless times.

For the purpose of this newsletter, though, the most important thing about them is that every year each of them holds a giant media festival, the International Journalism Festival in Perugia and advertising’s Cannes Lions Festival of Creativity.

Despite the fact that the two festivals, on content creation and content monetization, are linked and should have a lot of crossover, only a small group goes to both.

I have been one of them for the last few years, and I think that pulling out the values of each, and pointing out their shared misunderstanding of this moment, is particularly important when the map of the media and internet landscape is being redrawn on a daily basis, flooded with content. People in both guilds are responsible for deciding what the next map looks like and who gets to draw it.

The Rebooting’s Brian Morrissey, one of the others at both festivals this year, said in his Perugia session that the journalists there are part of a “guild.” So in the age of agents what makes these guilds worth joining and what do they build to survive?

Forts and markets

The defining physical feature of Perugia is its hilltop, crowned by a massive fortress built once upon a time by a pope, where you can survey the rolling lands of Umbria below you. Perugia is in fact exceedingly difficult to get to from most places, which fits with a vibe of being secluded and separate from the world at large. I love going to Perugia each year for the community, but there is occasionally the feeling of ascending the mountaintop to hear the wise high priests of capital-J Journalism speak.

Cannes has also had its fortresses (including a role in Man in the Iron Mask) but its defining feature is the beach and the Boulevard de la Croisette that runs along it, which becomes a long line of pavilions. That beachfront setting and harbor is also part of why Cannes and its surrounding islands have been repeatedly and predictably attacked, from medieval pirates to more modern wars.

Nowadays it is invaded by multiple festivals a year, with the Cannes Film Festival, Marche du Film and Cannes Lions always bringing to town whatever the hot new thing is. Cannes Lions and its side events, which are for the advertising and increasingly the AdTech industry, is a particular meta-exercise in influence and grabbing attention, as that is both the strategic aim of the industry and the tactical imperative of the event.

While Perugia is about adherence to a set of sacrosanct values amidst the constant refrain that we are in a “moment of transition,” Cannes Lions takes transition as a given and is a scramble to prove that your hot new shiny thing is the definitive hot new shiny thing, and everything else should be chucked overboard.

Meet the new boss, same as the old boss?

The hot new thing this year of course is a media environment mediated by AI agents, which impacts both the Cannes people and the Perugia people. For the former, it is one more attraction in the carnival of buying and selling, and OpenAI’s advertising arm will surely have a coming out party at the festival this year. At the same time for the Perugia people, the LLM companies were viewed at a distance, another foreign power out there to be managed and negotiated with, the same way search and social were.

The guild at Perugia has a shared sense of purpose, and I think, despite a differing purpose, the advertising folks, descendants of the “Mad Men” guild of yesteryear, do as well. Yet both groups are looking at “AI” as another in a series of challenges for their profession. But this framing misses the boat, or rather, the sea change.

The most interesting development of the entire AI cycle so far is not the growth of OpenAI or Anthropic as a platform, but the mass adoption of agents on a personal level. The millions of people who are Claude coding their own software are still a bit geeky now, but increasingly this will spread to everyone who can fork over $20 a month or less.

The biggest players such as OpenAI and Anthropic are desperately trying to centralize all of this experience into their own tools to justify enormous valuations that depend on having some sort of monopoly.

But ultimately the emergence of language technology is going to be a great decentralizer, with models as good as today’s frontier models running locally on your laptop. You will have an agent that goes and searches all the relevant MCP servers, coming back with what you want to see and building how you want to see it, without the need for a pageRank algorithm or black box being mined for greatest engagement and profit.

Decentralization like this is a big question for the guilds because the work of both creating and monetizing can be done by anyone, which is a threat to the entire concept of guilds. The “golden age” of traditional journalism had a monopoly on information in a certain area, and the advertising agencies had a hold on reach when they were the ones negotiating spots in the big papers and on the three TV channels that existed. Journalism’s monopoly on information has already eroded, and its grasp of defining what journalism even *is* seems to be slipping amid a tide of creators.

Now advertising’s control over the pipes, such as big bad programmatic systems, will also morph into something else as agents mean budgets don’t need to pass through centralized platforms taking their toll. Money will pass not only B2B or B2C but X2X.

Islands in the (bid)stream, that is what we are

This is a change of the way that media is created, received and monetized that is probably on par with the magnitude of the World Wide Web. To me, agentic AI is not another invader coming to challenge the fortress of Perugia and the market bazar of Cannes, but something that means the guilds of Perugia and Cannes are a much looser concept. The map that has marked big bold dots for monuments to the publishing and advertising industries will look more like a million little islands, floating in the sea of content.

The next question is how do we keep the good parts of both of these guilds while jettisoning the bad. In a conversation in Perugia, I chatted with a researcher who agrees the shifts are similarly seismic, and wanted to preserve the values of journalism in new systems. Fitting with our dusty old European town theme, it is a bit like monks trying to preserve literature and science in abbeys during the dark ages. In a bit more of a sci-fi direction, it is a bit like Isaac Asimov’s The Foundation, where Hari Seldon tries to lessen the amount of chaos between collapse of the old empire and the beginning of something new.

The ad people are also trying to preserve the creativity and craftiness that helped them reach people, with reach and resonance being some of their core values. The glory days of a creative genius like Don Draper conjuring up the next great commercial are long gone, but the job of advertising is even less glamorous if it consists of throwing money at one of five machines and sitting on your hands. Journalists were disintermediated, but advertising pros may be as well if there is always an agent sitting between their message and the human being they want to reach.

As we can tell from how the literal physical infrastructure of Perugia and Cannes helped shape what they became, the future of media depends on what we build and making sure that it still contains the values from the guilds. The era of hyper-scaled platforms had a certain set of in-built values that led to both good and bad. Publishers, brands and individuals are going to be a group of scattered islands, and we will need to build the bridges, dykes and ports that help each contribute to a vibrant archipelago.

From Pipes to Ports

As I mentioned earlier, I think that connecting different nodes around protocols such as MCP (or AdCP for advertising) is a step that can help preserve a new decentralized network of creators and brands from being dominated by a single platform. If each of us and our work is an island, these protocols create ports to export whatever we want to send out (our content) and receive as well, as long as we have at least some limited amount of spare boat capacity (a locally hosted open source model, for example).

While both publishers and advertisers are now thinking about big players such as OpenAI and Anthropic, there is nothing stopping them connecting to each other in mini alliances. Why has the New York Times not thrown its content up on an MCP server for all its subscribers who use Claude Code? Why haven’t groups of papers done the same thing collectively?

To belabor historical analogies even further, there does not need to be a monopolistic Dutch East India Company coordinating trade and manipulating prices to make exchange work when the reason that this technology has taken off is because it is accessible to pretty much everyone. We need to have reliable data (and Overtone is part of that) but the technology itself and its ability to write software means that anyone can build their own port.

As part of thinking outside big centralized models, I made a version of what I think one of these ports looks like. It provides information about news based on sources available via RSS, provides a content intelligence layer of structured metadata that is Overtone’s stock and trade, and provides specific functions where you and your agent can use the data to not just receive news, but analyze it in charts, graphs etc.

On the flip side, it also includes slots for advertising, including letting a user control what they see, so that brands can reach a receptive (and consenting!) audience in moments where their message makes sense, and the contributing publishers to that moment can be compensated for the work of their journalism.

LEARN MORE ABOUT THE SKILL

This is just one experiment, but I think that as more individuals and small companies access the web agentically, not just through a model or platform, we can create structures on top of this that enable even more sophisticated forms of connection, such as individuals joining together to form a social experience, and publishers joining together in impromptu alliances for sophisticated monetization.

The world of agentic AI is a world where anybody can have agency, and it is our responsibility to seize that agency to build the world we want, not just for cute demos but for infrastructure. We can take the ideas of Perugia and Cannes and build them into the map that is currently being drawn.

Paradoxically, the way to make the internet more human may be to have, at particular points, fewer humans involved.

All corners of the online content ecosystem are now immersing themselves in what it means to have systems operated by agents, which use language models in addition to other tools (web searches, APIs, calculators) in order to accomplish tasks.

These agents will impact many layers of knowledge work, including things like how people get their news. Our CEO Philip Allin has written his thoughts on how this moment actually creates an opportunity for journalism to restate its value not just as the writer of the first draft of history, but also in educating machines that assess that draft about events and why they are meaningful. Here he is on “liquid content” and the value that underlies it.

Read Philip's Piece

Phenomena like liquid content are part of a larger movement: the internet moving away from pages towards something more fluid. In many ways this is uncertain but it can also be tremendously beneficial in moving past some of the problems with pages, which led to an overemphasis on keywords within a page and clicks onto it.

An internet of moments

But what comes after? Overtone sees the internet moving towards a system of moments. Moments are events, like concerts and elections and business updates, that bring communities together across different chats, channels and geographies.

This can sound ephemeral, but it can also be tied to something as concrete as dollars and cents, where moments discussed by many publishers can be monetized at scale. Today we are announcing that our moments understanding is now available agentically. Those in the advertising industry will definitely have heard of the AdCP protocol and AgenticAdvertising.org, but those outside of ads should also take a look to understand how the publishing business can connect to the outside world of brands and agents and advertisers in a more nuanced way than the traditional pipes.

Previously, running an advertisement on major publications required large amounts of set up. Here’s a new, simpler workflow:

• Let’s say I am an advertiser selling cruises to Norway,

• My agent talks to Overtone’s agent about real-time moments that are happening, relevant to Norway, such as discussions of the country’s performance in the Olympics or the movie Sentimental Value.

• My agent then and agrees with dozens of news outlets about a fair price to “sponsor” that moment, all almost instantly.

In some ways it is surprising that the business/tech side of content is the one making the charge towards a more nuanced, human way of operating, but actually it makes all the sense in the world. Much of the inhumanity that has been crushing journalists and content creators over the last decades is because of pipelines and procedures. Now those processes are changing, in many places avoiding people who have been given bad incentives for keeping systems in place.

If you are a Claude Code or Codex power user and want to play with Overtone’s agent, drop us a line.

I am Agentic

If you are a publisher who is interested in connecting into this world of agents, email us here.

Make Me Agentic

Omnichannel Moments and GEO

As we alluded to in our last newsletter showing off our moment capabilities, one of the beautiful things about this sort of work is that it can understand content as a human would *across channels* as someone goes from an article to YouTube to a podcast, etc.

This level of understanding is also important for the emerging field of GEO ( Generative Engine Optimization), where people know how their brand and voice show up in chatbots like ChatGPT.

Systems like GPT are very much black boxes, but they also sit downstream of the actual world and the journalism that tells us what is happening in it.

Overtone monitored a month and a half of data for both articles and LLM responses on subjects as diverse as the UK Labour Party, Coca-Cola, Manchester United and Bad Bunny. This allows us to extend our “moments” to chats, so that someone could bundle together articles with YouTube with someone on GPT. See below for moments that starts in articles before impacting GPT later, with discontent about the Labour Party leading to an uptick in sadness in chat responses, and happy news about a new stadium and transfers doing the same thing for Manchester United.

Beyond that, because Overtone has deep understanding of the articles that make up a media moment, we can understand why some moments (such as their types and tones) make the jump from articles and videos and podcasts into the responses of a chatbot (and how they differ from one to another). Get in touch to learn more.

I am Interested in Chats

At the intersection of online content creation and advertising, there is often a fizz of energy. It’s the noise from a trillion-dollar machine that is relentlessly trying to find better ways to produce work, or profit from it.

Right now, the buzz is around liquid content. This is the idea of having little bits of information (“It’s raining in Tokyo”, “Acme Corp is up 4%”) and allowing them to flow across different mediums and in different shapes and sizes, without losing their context or their value. It’s useful to dig in here for two reasons. First, it helps us understand how AI creates content, so it gives us a look to the immediate future of the content industry. Second, I believe it will have a radical impact on content businesses (publishers, platforms, individuals, brands and so on). Handled well, it can benefit everyone.

Liquid content is a direction, not a trend

The modern publisher is fighting a losing battle on three fronts: platforms (who determine the access), advertisers (who own the spend), and what we might loosely call AI (which increasingly manages the interaction).

Underlying all this is a more exhausting battle against human complacency. Faced with a choice between apparently good enough content –for free, just accept some cookies or a privacy violation and go– and clearly well crafted content, most users, most of the time, will take the path of least intellectual resistance. The Fourth Estate is fighting the ‘effort paywall’.

Take this recent Rebooting piece, which makes the case for moving from output-based to usefulness-based journalism. This principle helped Dmitry Shishkin triple audience growth at the BBC. A key point in the article is the suggestion that newsrooms can easily get distracted by ‘shiny object’ formats (VR, or specific app features) without actually supporting the core mission.

In the piece, Shishkin asks for examples of publishers pushing out liquid content. This question sidesteps an important point. Publishers are late to the liquid game. AI, mostly in the form of LLM-powered chatbots, are designed to work in liquid form. This is because of the nearly direct interaction between the consumer and the content itself. Information is thrown into a vast tank, and as far as AI is concerned, is no longer tied to a fixed format. Instead, it exists as a fluid asset that can be shaped to an individual’s consumption patterns in real-time. You want the weather report as a video? Here it is. Your friend wants it as a table and some bullet points? Voila. The same basic information, very different formats.

Turning the page

So how do creators and owners, analysts and advertisers (and let’s not forget the actual readers!) deal with a paradigm shift of such unheard proportions, which will likely mean leaving ideas and the economics of things like “the page” behind? Before taking a step forward, it can be helpful to take a few back. Here’s a rough timeline of how we got here. Ping me in a few years if the prediction didn’t work out.

• 2000: Mass Adoption – Internet growth and digital choice allow people to navigate the web.

  • The internet is a mall.

• 2005: Social Media – Information flows through personal and professional networks built on human trust.

  • The internet is a business hub.

• 2010: Pervasive Feeds – Passive consumption eases to the fore, as people rely on networks built on implicit trust.

  • The internet is a party.

• 2015: Personalized Realities – Echo chambers become public, as highly personalized world-views shape real-world events.

  • The internet is a filter.

• 2020: Machine Learning – Transformer-based architectures quietly reshape how we experience our understanding of global events.

  • The internet is a mirror.

• 2025: Customization Rules – Chatbots and vibe coding allow us to create strongly personalized experiences around information.

  • The internet is a feeling.

• 2030: Hyper-Embedded Customization – Your online (and most likely offline too) life may be entirely different from anyone else’s, fully realized through the flow of a liquid internet.

  • The internet is a first-person game show.

Financing the future

So how do professional writers, those who really care about their audience, fit in? They become minters of value. For better or worse, online content is now a marketplace. Consider the finance industry, which is a roughly 30x larger market than the content world. Part of its enduring success is its willingness to treat value as fluid and transactional. It is, in a word, liquid. What if we force ourselves to think no longer of strict pages or articles, but in useful elements, such as the data behind a weather report, or a public figure’s recent quote, or today’s Acme Corp stock price compared with last week? When content is dissolved into verifiable and modular elements, it creates information flows that can be shared, streamed and measured around the world at near-zero marginal cost.

Then the role of the creator –the human, don’t forget– transforms into a reviewer of what’s going on. This arbiter of relevance can then be paid, according to market demand. This is unlikely to be an immediately popular idea, certainly not with professional writers and creators. But it is the most logical way for them to be valued for their talents. Similar to media outlets, these people exist as a medium between a reader and the world, acting as an authority on recognizing relevance.

So this is a solution to the content polycrisis. In a world of endless AI synthesis, real value resides in whatever is new (and true). It should make sense: AI, for all its charms, cannot know what is going on at any given moment. We, people, are the source of the raw and verified account of what is actually happening. The moments that matter.

Pipes and pages

The current content industry processes a trillion dollars through a system built around the idea of pages, posts, and platforms. I think this is why there is a kind of industry stasis. The systems we have aren’t designed for the content experience we demand, but no one has yet cracked the puzzle of re-engineering the massive machine to ensure liquid content powers business liquidity.

In this agentic future, the ‘page’ evaporates. Journalists and publishers must shapeshift into the marshalls of a liquid world, providing the information flow with legitimacy. Assuming we create infrastructure and an eco-system to match, humans will not only have a key part in the future, we will be rewarded for it.

From the First Draft of History to the First Draft of the Future

Journalism was traditionally the ‘first draft of history’. Properly understood, integrated and managed, liquid content provides the opportunity to flip this around. Content is no longer just a record of the past. It can serve as a record of what is happening now, and feed into what is next.

In this world, people craft raw material upon which all subsequent versions of history are built. We can then return a significant percentage of revenue to the person or people who invested the time to record something of value. That something starts with a fact and adds perspective, experience, humour, or comparison: something personable, something that explains why this story or picture matters. In a world of fluid information, you either provide the source or you become the sediment. It is time to go with the flow.

It’s got singing and dancing, winning and losing, multiple languages. It’s politics, sports, entertainment all rolled into one.

The Super Bowl, the championship of the National Football League, is one of the highest concentrations of attention throughout the year in terms of sheer attention tonnage.

This makes it a yearly Mecca for advertising, but not without chaos. Perhaps no event better embodies what Overtone calls the flood of content. It is colourful, loud, omnichannel, concentrated and dispersed, the event itself and the backlash to the backlash to the thing that’s just begun. It can feel overwhelming, like the senselessness of putting *everything* on a bagel.

As discussed in our last newsletter, this can be incredibly alienating. But part of the mission of Overtone is to help people and companies understand the internet in a way so that they can make sense of it and contribute by putting their messages out there. So we decided to bring in as much content about the Super Bowl as possible from as many different channels.

This included live transcription of the Super Bowl as it was broadcast, images of the broadcast every 5 seconds, thousands of articles from traditional media outlets, dozens of YouTube channels, broadcasts the days after, and regular questioning of GPT about what it thinks is going on (more on this in another newsletter). All of this is timestamped into a real-time multidimensional, ever-growing map of what happened.

Insights: Understanding the moment

Maybe, for example, I am a travel company who wants to get the word out about deals to the Caribbean. Fans of halftime show artist Bad Bunny are the sort of people I want to reach, but I have a big problem. Bad Bunny is mixed in with the cacophony of the Super Bowl, with his name appearing everywhere from TV listings to transcripts to tweets.

What Overtone allows is looking at the narratives around Bad Bunny, before, during and after the Super Bowl, including which other concepts and tones are most tightly tied to him for narratives that have a particular emotion I may be interested in. You can then discover, automatically without needing to watch anything, how narratives around a subject are changing.

• Avoid the noise: We can steer clear of articles focusing on politics or divisive figures such as Kid Rock.

• Find the fun: Our model found that surprise guest appearances, by Lady Gaga and Cardi B. among others, created adventure and joy narratives, which is exactly where this travel brand wants to be.

Activation: Be in the moment

Understanding what’s happening online is one thing: acting on it is a different ballgame. Because Overtone uses semantic, emotional and image based signals (and zero personal data, ever) you can place ads next to exact right content on any channel.

TV Images: Overtone’s system analysed almost 4000 images from the Super Bowl broadcast including the halftime show. Of all those more than 5 hours of images, the model for the travel campaign said that the images that were the most resonant were when Bad Bunny put on a wedding. He danced with Lady Gaga and others to his song Baile Inolvidable as the party swirled around (6:30 and onward here). Other top moments included fireworks and another big dance number.

Articles : Of course the moment of the Bad Bunny wedding did not live only on NBC and other broadcasters. Moments as they are understood by Overtone become narratives that spread and change as they do. Our Bad Bunny happy travel moment has a peak during the broadcast itself, but also bumps afterwards such as reporting that explains the wedding on stage was in fact a real wedding.

YouTube: The moment continues on as it moves from the written word to formats that take longer to produce, such as YouTube videos. You can see in the chart below the moment continuing with a wave the following day, Here also Overtone’s ability to highlight specific narratives is important, with some news creators playing up the culture war angle you may want to avoid, and others focusing on the wedding and the celebration of Latin America that is more aligned with the travel campaign’s ideas. As this content is included in the analysis, the model also updates itself to take in new information about what might be relevant as time moves forward.

Omnichannel: CTV & LLMs: This narrative understanding can go anywhere where the content is available. For example, when two days after the Super Bowl a local New York NBC station did a segment on a dancer who dressed up as a bush during the show, it would be possible to use Overtone to run your aligned ad campaign in the next available Connected TV slot on the streaming service Peacock.

Similarly, the world is beginning to understand how advertising on LLM interfaces might work. Overtone measured how one LLM responded to questions about Bad Bunny submitted every day, and tracked the answers. Unsurprisingly, there was a boost in happiness the day after the Super Bowl. If I am an ad agency I can choose when I would want to have my ads next to a subject being mentioned by GPT, and when I might want to avoid it.

Sponsoring the moment

Overtone can do all of this monitoring not just because it is interesting to know all of the different nuances of how narratives form and change. Ultimately we believe that this level of insight gets into a more human, more artful way of understanding the world online, which has traditionally been dominated by blunt metrics such as clicks and keywords without knowing the full context of the cultural moment.

Using Overtone’s moments, you can track the narrative as it spreads and make sure your message is there too, essentially sponsoring the moment like you would sponsor a match or conference (or wedding) in the real, physical world.

This was not possible beforehand, but now can be done at the speed and scale of the internet, making sense of the chaos out there by putting it on a map, and watching that map change over time.

Overtone is a machine that tries to understand moments like a human, but can take in the everything everywhere, all at once.

If you would like to see our Super Bowl data or discuss a big upcoming event (like the FIFA World Cup) get in touch at business@overtone.ai

“Most of the words you’ve ever read were written by a human, but that could soon change.”

That was how I began an essay almost exactly five years ago, which I titled “The internet is not ready for the flood of AI-generated text.” I wrote it after the first few months of GPT-3 (before “AI” was cool) and believe that the main thesis has held up pretty well through all of the hype and reality that followed. The systems that shape information environments care more about wringing every last ounce of engagement than about people or their best interests, and many have been overrun.

Now, the flood is all around us. A report from the SEO research company Graphite estimated that as of earlier this year, more than half of all articles online were created by generative AI tools, beating out those created by humans. Users of platforms such as X and Instagram do not need me to tell you that there has been a burst in AI bot accounts or AI influencers and actresses such as Tilly Norwood.

The result, as predicted, is what I called the “infinite Times Square,” where any area not locked off is a mix of real and not real, aggressively competing for attention and trying to seem “authentic.” We see this in everything from bots to an Italian philosopher named Andrea Colamedici writing a book about the dangers of AI “Hypnocracy” with the help of AI and then releasing it under the pseudonym of a fake Chinese academic.

Dangers from this have included the ones that I mentioned five years ago, with some calling the flood of AI slop as an example of “enshittification”, some looking at its impact on misinformation and disinformation and recent research suggesting that smaller publishers and creators are like small islands going underwater as a narrow group of larger outlets are disproportionately cited in AI answers.

This is where I thought, and still think, that new forms of data, like those Overtone was built around, are key to making the internet more human again. The previous essay was a jumping off point for the founding of the company and we now offer content understanding data for everything from the level of journalistic depth, the type of article, the concepts mentioned, the emotional tones, potential misinformation signals and more.

All of this can be combined into customisable models our clients use to better understand the internet around them and find the most relevant articles (as well as audio and now images) to support. The first step in dealing with a flooded landscape is to have a map. Solid data is that map.

Solipsism, communities and ChatGPT with ads

But now that the flood is here, we can see worries beyond the paradigm of what we cared about as issues five years ago, ones that emerge when technologies are widely used.

One of the changes seeing widespread adoption for the first time is mass personalization of content, through mechanisms like chat interface responses. Levels of personalization have existed before, through algorithmic recommendations on social platforms, for example.

However, the particularity of chat responses, coupled with the mixture of real and unreal seen in episodes like the Colamedici example, leads to something else. Alienation from an external world that you cannot understand as real or not real can lead to an isolation so profound that it is akin to solipsism, or a skepticism about everything in the external world beyond your own self, a la the brain in a vat scenario. One of the more disturbing threads around ChatGPT specifically in recent months has been its connection to suicide.

The current and future levels of personalization differ from algorithmic recommendation on Facebook, for example, because of the memory and specificity of the experience but also the sheer size of content when used at scale on a societal level. One of the dangers is not just the chatbot responses themselves but because the information ecosystem now has so much content, the percentage of all that content that one individual can see is getting smaller and smaller.

These low percentages mean that the percentage of content you’ve viewed and that someone else also has, so that you can talk and discuss it in your communities, will decrease and decrease. The information ecosystem will become further and further fragmented so that we are all essentially islands.

This fragmentation is on the verge of being deeply enmeshed with the way that the internet is funded, as OpenAI has been actively recruiting for advertising-focused roles for a while and CEO Sam Altman has said he is not against adding ads to the platform.

But supporting the hyper-fragmentation of our information environment with advertising budgets is not good for the information environment or advertisers themselves.

Over the last five years, Overtone has worked with actors all over the information ecosystem, from publishers to advertisers, and the thread that unites the good actors, and differentiates them from the bad actors in every category is a focus and care for the communities they serve and work with.

That is why Overtone has seized on the idea of “cultural moments”, using our models to understand what tones, types and topics make a community tick and following them through time as different events pique their interest and bring them together in discussion, rather than trying to gobble up all the possible personal data such as whether an individual bought a toilet seat last week. Brands that want to reach their audiences can be there with the communities they serve, essentially sponsoring moments online the same way they sponsor in-life events, finding places in the flood where people have gathered together for something real.

Going forward

I increasingly believe that moments, real-time discussions accessible to all without control by a centralized algorithm distributing them, should be the basis for how the internet operates and is funded. Data about moments and the content that make them up is essential here as well, and can serve as a common standard of the characteristics of different moments so that they can be understood at a glance.

Data about content can function not only as a map, but a passport, a face to meet the faces that an article meets as it is read online and becomes part of larger cultural waves. This is scalable from angry outbursts about political corruption in your local town to global outpourings of praise from wildly different corners for Rosalía’s new album (very much having a moment).

As the future of journalism is reimagined by thinkers such as David Caswell, I believe that the right data can also help serve the underlying mission and purpose of helping everyone online understand the world around them at scale and in the granular detail of linguistic nuance down to the paragraph level.

Looking forward to the next five years of artificial intelligence, I see that our current moment is still very much focused on trying to get AI to accelerate and optimize processes that have existed before, with questionable results. But the biggest opportunity to use AI for good isn’t in streamlining the same flawed systems that we had before, but in achieving things that would never have been possible without it.

There is an internet waiting to be born that understands the humanity and creativity of culture in real-time and can connect users in communities without control by a centralized algorithm. Building artificial intelligence systems to help create the data that powers these systems and enable larger scale understanding of nuance remains the best use case for AI.

Overtone is excited to announce that we are now in the final phases of testing our models for content understanding not just on words and transcripts, but on images.

We are not the first technology to extract insight from images. Others use approaches like object recognition for weeding out pornography or recognizing a company’s logo in an image. However, our progress into images means we can bring something new to the world, our focus on creativity. One of Overtone’s goals is to elevate the way that ads slots are sold online by using AI to bring creative understanding.

In the media industry, focusing just on the objects present in a photo plays into the same trap as keywords within text. They were a useful way of sorting through content years ago, but two articles on the same topic can come across in very different ways.

Current technology such as Overtone’s enables us to achieve a more narrative understanding of what a human being is experiencing as they take in someone’s work, be it a news article, podcast or photograph.

Those moving and interacting with that content around the web, from the newspaper putting it in a newsletter to the advertiser placing an announcement next to it, can understand the real context of what is happening and work to make better human experiences.

Creative Understanding

Take for example, this image of HAL 9000, the evil artificial intelligence at the heart of Stanley Kubrick’s 2001: A Space Odyssey. If you are looking at the objects within it, it is a picture of a lens, or of a computer.

But what the viewer knows, and what Kubrick knows when he puts the image in his film, is that it is terrifying.

Overtone found that it had one of the highest probabilities of being “fearful.” (By contrast, of the movie images we analyzed, a prancing Julie Andrews in The Sound of Music was one of the most probable for “happy.”)

Our system can tell us this, not because it has a deep knowledge of 1960s cinema, but because it can analyze the image itself for its emotional resonance. Understanding the essence of a photograph is actually a concept that has been discussed in photography for a long time, such as Henri Cartier-Bresson’s “Decisive Moment”.

Be a Champion

If this sounds pretentious and arty, it’s because compared to a lot of the ways that content is currently distributed and monetized online, it is. But it also has immediate practical applications making sure that ad placements better align with what an audience is experiencing as they consume content.

In this spirit of quality, Overtone looked at several thousand articles from the first week of league phase action in the Champions League, the pan-European football competition whose reason for being is organising high quality matches with “die Besten, les grand equipes, the champions!”

If I work for Heineken or Playstation or Takeaway.com or any of the other sponsors of the Champions League, it makes sense for me to advertise on articles that people are reading about the matches, but there can be too many to choose from. I can target a keyword like “Arsenal” but this leads to articles with many different narratives.

Below is one of the happiest articles we found in the sample, with Leandro Trossard and Gabriel Martinelli smiling after they came on as substitutes and helped the London side beat Athletic Bilbao 2-0. Meanwhile with very similar keywords (Arsenal, Athletic, Champions League) the advertisement could appear next to an image of Arsenal’s captain Martin Ødegaard on the ground after sustaining an injury, which we labelled as one of the saddest images in the set and is not exactly something you would cheers to with a beer. Arsenal is also mentioned in an image of the league table, which we rate as purely “informational.”

A More Human Internet

Overtone’s mission is to give human beings clarity on content online. We are excited to continue this work by exploring how a deeper, human understanding of how images resonate with people. This will allow us to move beyond rudimentary tools like keywords and elevate the granular data of programmatic advertising into something more creative.

Next steps include integrating image understanding into existing products such as Community Lens, that can understand what images speak to a certain community, and Culture Pulse, to understand how those communities grow and discuss over time. If you would like to be part of this, send us an email and let us know what you are thinking.

Send us an email

We believe that this insight will help both the advertisers, who get to connect to audiences more authentically, and the publishers who put out quality work that can be drowned out in a world of AI slop. Especially with language model generation, we should work to highlight the work of photographers who create *the* image of an event, rather than just *an* image.

A more creative internet is one that understands the value of humans making art and journalism, sharing it with other humans. That, to us, is the bigger picture.

Overtone would like to thank students Mila Pretolani and Bayli Wolfe for their contributions to the image analysis framework this summer.

Christopher and Philip were happy to be back in the south of France for Cannes Lions this year. Find below Christopher’s reflection about the importance of the news industry in grounding advertising amid a sea of changes.

Read the Post

Overtone Community Lens

Online we are often alone in a crowd. You scroll through the “for you” feed on a social platform or other recommendations designed to have you engage and keep scrolling.

The monetization aspect of the internet also works similarly, with systems designed to find out where you were last week, what your identity is and whether you are likely to buy an object like a toilet seat or not. If you are, you may receive an advertisement for one. Or ten.

Individuals may buy things, but it’s the communities they are part of, families, jobs, fandoms, that really are the focus of their lives. Brands, from soft drinks to software, realize this and increasingly are trying to connect to those communities as a place to do business. Overtone believes in maintaining people’s privacy and thinks that the best way to understand communities online is through the content they read, watch and listen to.

Content-based Audiences

Let’s take for example the case of an advertiser who is telling people about an upcoming show, in this case Giuseppe Verdi’s Il Trovatore, which is on at the Royal Ballet & Opera in London. As that advertiser, I want to reach people who would be interested in coming to this opera, which is dark and dramatic.

I can go looking for some sort of data that spotlights a dark and dramatic person, or search for individuals who used their credit card on a ticketing website in the last six months, but there is another way that is more privacy focused and more resonant with my ad banner (in the image below).

Now that technology can have a fine-grained, more human understanding of content, I can use a tool such as Overtone to find the places where my audience is already in the mood for potentially going to the opera.

After all, my potential customer may be a banker in the City of London or a father of three somewhere in Kent, but if they are reading about stock prices or updates to the local school schedule, they are essentially a different person at that moment from the opera lover I want them to be. The article one is looking at makes the audience.

In the example below you can see two articles from the same outlet about entertainment, which both mention keywords like “drama.” But Overtone’s tech (having analyzed the brief) can recommend that your ad appears on the article about theatre and screen icon Richard Burton, but not on one about reality TV. It can pick up the subtleties of tone, the way a human would, to make a recommendation.

What makes a community

Because the way that Overtone’s tech works involves fine-grained understanding of things like the emotional tone, level of opinion and type of journalism, this can also help people trying to reach an audience understand exactly what makes that audience tick.

In the example above, the model that was built from Overtone’s signals to reach potential opera-goers can be used to evaluate thousands of articles from around the world, and give them a score for the opera campaign. But, in a big difference from asking an LLM to do this, for example, it can also use concerete data to show us why it is making that decision and let humans tinker with decisions they don’t like.

So for the Il Trovatore campaign, we can see that model is more likely to rate in-depth articles as the sort of articles our opera goers read, or articles that mention “composers” or articles that mention the TV series Succession, which while set in contemporary New York, also has a high level of drama. Similarly, if an article is a listicle or about Sabrina Carpenter, the people we want to reach are probably not in the group reading it.

Communities in Real Time

Reaching communities through what they read is particularly powerful when you can understand that community over time, as Overtone discussed in our last newsletter. When a community around a TV show, sports team, concert or city is reading articles that speak to them as a group, they are no longer just isolated individuals, but reachable as a collective.

While the hyper-connected internet can actually be a very isolating place, understanding where readers are in their lives at that moment, from checking football scores to ordering diapers, can help brands speak more directly to their potential customers and support the content that community reads. Using AI can make things more human.

The Cannes Festival of Lions, and all that surrounds it, is often a blur, with tens of thousands of people, dozens and dozens of stages, and about a million crafted messages all competing to become the message that resonates and forms a hot new trend.

With the flood of attention around AI or connected TV advertising or content creators all buzzy and “new”, a focus on news or journalism can seem like yesterday’s news.

But at Cannes and in other places in the media ecosystem, new narratives are taking place around news that show it finding its place in the chaos of the broader ecosystem.

From safe to special

The discussion of news always begins at “brand safety” or concerns about appearing next to potentially controversial content that a brand views as damaging to their ability to market whichever product they are trying to sell. Estimates say that close to half of news content experiences blocks from receiving money from advertisers based on these concerns, which largely occur because they mention keywords like a politician’s name or “COVID.”

What started as brands not wanting to promote conspiracy theories and the like has now morphed into what one attendee of the Future of News breakfast in Cannes called the “secret assassination squad” of news.

There has been good research about how the alleged negative association of brands advertising on news is actually not real, though there are now efforts to change the narrative around news in a positive direction. Not only is news not unsafe, but valuable.

Work is ongoing to shape that exact message to advertisers and back it up with data, though I found it interesting that the ad agency behind the Future of News breakfast, Stagwell, also places a major emphasis on sports, with their “Sports Beach” venue in Cannes filled with celebrities from basketball, running and football.

Big brands and their agencies are very much interested in being with communities of sports fans as they go through the ups and downs of their teams’ successes and failures. One of the most common refrains at Cannes this year was “connecting to culture” such as sports and entertainment.

In terms of the timeliness of connecting to that culture, I can think of few better ways than by doing so through news. There may be a lot of chaos happening online everyday as billions of people post, like, comment and read, but news turns that into a “first draft of history” that brands can understand and interact with. News has similar communities to sports fans, but instead of following their team they are looking at the town, city, industry or other group they care about.


Disruptive technology for good

Beyond the folks at Stagwell, one of the other people pushing news back into prominence is RTL Group, the large European conglomerate that owns dozens of websites, radio and TV stations.

One of their main messages was that news is not only safe, but particularly safe because many RTL brands are known. In some ways, it may be tempting to view major media groups like this as nostalgic for the days when they were the biggest show in town. It reminds me of 30 Rock, when one of the programming priorities for NBC was “Make it 1997 Again through Science or Magic.”

Technology won’t eliminate all of the competitors for attention that a major group has, but it can help them solve another of their problems, which is that ad buyers don’t want to have to deal with many smaller players, and often prefer to be able to buy everything in a programmatic or platform package.

“There is overhype on AI, but we are bringing AI out of the bottle to solve real life problems which are sometimes very specific to the industry,” said Thomas Servatius, co-CEO of smartclip and leader of advertising technology at RTL.

Servatius explained that he has embedded agentic AI on the ad server, so that it can better understand requests coming in from unknown ad buyers, for example by looking at the domain that the campaign wants to direct readers to.

He also sees shifts away from the keyword-based approaches that have been problematic for news producers for so long.

“It’s a pity. We invest a lot in creating valuable news and journalistic investigations. And it's hard to monetize, yes. There I think the industry starts to be better because we are now have developed more and better technology, technologies to pass on content signals, to standardize content signals.”

Standardizing content signals may be the next step in making sure that the value of news, having a steady, experienced finger on the pulse of culture, can translate into the nuts and bolts of the advertising system.

The signals that a brand uses for understanding one publisher can then translate into understanding of a whole market or region of the world, and for agencies that are working with multiple clients across multiple markets, help them build their own learning and best practices. This is not possible with pure LLM prompting approaches, where you can get a different answer to the question “is this good for my ad placement” depending on the time of day, and individual models remain in silos without a concrete, replicable understanding of the world.

The first draft of history connects us all

All in all the flurry and tumult of Cannes are a good reminder on the value of standardization and transparency as AI tools and offerings spin up new versions. What is behind it all, connecting it to reality?

News, including coverage of big chaotic events like Cannes, has the advantage of being grounded in the real world, reporting on what is happening and what is trending – and then interpreting it for readers who trust the author.

In a world where more and more there is a flood of content and possibilities, brands should welcome the ability not just to be on solid and safe ground, but to be able to understand the culture they are connecting to in real time.

Overtone is thrilled to announce our partnership with PubMatic, a major platform for monetizing media through advertising.

Together, we're redefining the future of transparent, content-first audience curation. Overtone's contextual intelligence enables brands to connect their investments with the narratives, communities, and cultural moments that matter, while simultaneously creating more value for publishers across the open web. We are excited that people can use our data to drive ROI for advertisers while supporting quality journalism, culture and consumer trust.

Read the full release

Overtone and Culture Pulse

From the partnership announced above, you can see that journalism is important to advertisers, and Overtone is keen to help highlight the value.

But what exactly is that value? The role of journalism in preserving democracy often gets the attention, but more tangibly, news has the value of bringing together communities and helping them understand what is going on. The “media” is a medium between a person (or audience or potential customer) and the world.

Looking at it this way, the articles we read and the podcasts we listen to are like measuring the heartbeat of culture happening all around us. Being able to connect with audiences as they experience culture, from the Super Bowl to the previews before a movie, is valuable to advertisers. It is the focus of Overtone’s new product to help advertisers connect to and support news, Culture Pulse.

Coachella

Most advertising is currently done with the blunt instruments of keywords, but these do very little to actually understand the experience that the reader is having on their screen, wherever they are.

For a first example, pretend you are a company that wants to advertise around the annual Coachella music festival. It’s a target for food, drinks, or fancy new earbuds. As an advertiser, I can target the keyword “Coachella” or even “Coachella Valley Music and Arts Festival” but that still puts me on a bunch of articles that don’t really fit my audience, like this article from Forbes about a “Golfer’s Desert Oasis“. Presumably those golfers don’t need earbuds.

Overtone allows companies to target on a more nuanced understanding of articles, including the type of articles and their tones. So instead of just targeting a keyword, I can target concert reviews from the festival itself. Moreover, let’s say my product is tied to a certain spokesperson, like Lady Gaga Oreos. I don’t want to be on the places where Lady Gaga is a passing mention, but on the places where she is a superstar. Overtone allows you to target and be part of the cultural moment when hundreds of articles are raving about Gaga and her “Genius Commentary on Fame”.

In some ways, this approach mirrors how advertising works on television. When I buy a Super Bowl ad, I know exactly what kind of experience the audience is having — they’re likely gathered around their screens, emotionally invested, and culturally tuned in. That context allows me to craft a message that resonates. Often, the ad itself might feature an NFL player or reference the game directly. It’s that shared understanding of the content and moment that enables the brand to be in meaningful conversation with culture. That aspect is sorely missing from much of online advertising.

Subscribe now

Wimbledon

Speaking of sporting events, one of the benefits of running ads around cultural moments is being able to understand what is happening in a world that is unpredictable.

It is not difficult to imagine that Lady Gaga would get headlines around Coachella, given that she was one of the headliners. However, there is no way to know before an event like a tennis match who will come out on top and what the storylines will be. The articles written about it can help you understand what is going on without needing to sit and watch the match.

As a second example, consider last year’s Wimbledon Championships. For a brand like Uniqlo, where Novak Djokovic serves as a global ambassador, you want to be closely associated with his presence and achievements. In the chart below, which tracks sentiment over two weeks in July, you can see a spike in happiness as Djokovic celebrates his 350th Grand Slam singles match win.

But you can also see a burst of angry articles on Djokovic on July 9th. This marks a moment no brand could have predicted— the player slammed the Wimbledon crowds for cheering for his opponent, setting off a mini firestorm in the British press. If I’m Uniqlo, I might be comfortable with the controversy, but I’d likely prefer to be associated with Novak having a heartwarming moment with his daughter (another spike in “happy” signals from Overtone) rather than when he’s making headlines for complaints.

White Lotus

Sure, if I’m running an ad campaign, I want to keep an eye on my stars and steer clear when they’re involved in something I’d rather not be associated with. But the value of connecting to culture through news goes far beyond just managing risk.

With Overtone, you can uncover cultural moments as they unfold in real time. You never quite know what part of a TV show or news story will spark conversation at the digital water cooler the next day—but if you do catch it, you have the chance to reach your audience with something that feels fresh, relevant, and urgent, rather than just another piece of stale corporate messaging.

Take the show White Lotus, which aired its third season on HBO earlier this year and saw a major surge in viewership. If you’re a rival streamer like Disney+, looking to advertise your own content to that audience, you might know in advance that the show is set in Thailand or touches on themes like Buddhism. But what you don’t know is what cultural moments, memes, or inside jokes will emerge and take off. That’s the unpredictable magic of live cultural conversation—and the opportunity for marketers who are paying attention in real time.

With Overtone, you’d see that audiences had a huge reaction to a surprise appearance by actor Sam Rockwell, who was not one of the originally announced members of the cast, when both he and “cameo appearance” spiked joyful and humorous conversations around the show. Similarly, take something like the election of a new pope. You wouldn’t know the chosen name in advance, but the moment it’s announced, that name—and the newly chosen papal title—suddenly become culturally and emotionally significant. Overtone helps you spot these unpredictable but powerful moments as they happen.

Back to White Lotus, I also wouldn’t know in advance that a blender would become a major plot point in the season finale. But with Overtone, once that trend surfaces, I can act on it immediately in my own campaign—assuming I’m not deterred by the fact that “poison” began trending alongside it.

Wrap Up

Overtone has always championed the power of contextual data—using information from the surrounding environment, like an article’s content, to place ads, rather than relying on invasive personal data collection. Culture Pulse builds on that foundation, offering not just page-level context, but cultural context: insight into the broader moments audiences are living through, reacting to, and talking about in real time.

We look forward to working with more brands and publishers, connecting their audiences to the world in order to further break down the atomized nature of online ecosystems, where the only thing that matters is the click. We believe there’s a better path forward—one that treats people like people. By approaching culture through a more human lens, we can help make the internet feel less like it’s designed for machines, and more like it’s designed for us.

Announcements

Event season is upon us and Overtone will be out and about this spring and summer if you’d like to meet us in person.

1) Overtone will be back at the Cannes Lions Festival of Creativity in mid-June. Let us know if you’d like to grab a coffee or hear more about our work – or to tell us about yours.

2) Overtone has been selected as a finalist for ImpactDeal which highlights impactful, sustainable startups in Europe. Philip will be on stage in Turin in early June speaking about a project on sustainability.

With Contributions from Annie McCarter

The partnership sets a new standard for how digital advertising can deliver performance, relevance and responsibility – ensuring that advertising not only drives ROI but also fuels the open internet, professional journalism and the stories that shape our culture.

Overtone’s proprietary AI technology identifies the narrative context of digital content at the page and paragraph level in real time – going far beyond the capabilities of traditional brand safety filters.

In this partnership, Overtone’s technology is applied across the open internet inventory available via PubMatic’s platform. Inventory can be segmented to create brand-resonant contextual audiences, enabling advertisers to run campaigns in relevant, high-quality environments, including previously excluded inventory like trusted news content, while maintaining brand suitability.

“At PubMatic, we believe digital advertising is the lifeline of the open internet – supporting professional journalism, cultural storytelling, and the content that connects communities. This partnership with Overtone enables us to help brands show up in meaningful, contextually relevant environments that were previously out of reach due to blunt brand safety tools. By applying nuance and intelligence to inventory curation, we’re unlocking new value for advertisers while reinforcing our commitment to funding a diverse, accessible, and sustainable digital content ecosystem.”
— Emma Newman, Chief Revenue Officer, EMEA, PubMatic

“Advertisers are now demanding more than just performance and safety – they want meaningful connections with customers at scale. Together with PubMatic, we’re excited to meet that demand by introducing contextual resonance infrastructure directly into the global programmatic supply path. Narrative matters, and this partnership enables advertisers to identify environments that are not only brand safe and suitable, but also contextually aligned and emotionally resonant with their target audiences.”
– Philip Allin, CEO, Overtone

As the industry progresses towards a cookieless future, brands and publishers are in urgent need of scalable, privacy-first alternatives to audience targeting that don’t rely on third-party data. This collaboration directly addresses three of the most pressing challenges faced by digital advertisers today:

• Blunt brand safety tools often conflate risk with relevance. Overtone’s paragraph-level narrative analysis, against a range of dimensions, unlocks safe, brand-aligned, resonant media environments

• Overused off-the-shelf audience segments often lack freshness. Overtone identifies and maps contextually engaged communities in real time

• Cultural opportunities that help build brand value are frequently missed. Overtone enables real-time activation during live moments—across news, sports, entertainment, and industry event

This partnership enhances PubMatic’s efforts to deliver transparent, high-performance solutions while reinforcing the company’s #fundedbyadvertising mission: to support the content, journalism, and storytelling that keep the open internet vibrant and accessible.

About PubMatic

PubMatic (Nasdaq: PUBM) is an independent technology company maximizing customer value by delivering digital advertising’s supply chain of the future. PubMatic’s sell-side platform empowers the world’s leading digital content creators across the open internet to control access to their inventory and increase monetization by enabling marketers to drive return on investment and reach addressable audiences across ad formats and devices. Since 2006, its infrastructure-driven approach has allowed for the efficient processing and utilization of data in real time. By delivering scalable and flexible programmatic innovation it improves outcomes for customers while championing a vibrant and transparent digital advertising supply chain.

About Overtone

Overtone is a contextual intelligence company redefining how online content is understood, navigated, and activated. Its platform analyzes real-time digital news and media and generates transparent, signal-based insights into how stories resonate with audiences and align with brands— revealing not just what content says, but how it is perceived by people. Built for the advertising, publishing, and public relations industries, Overtone eliminates guesswork with verifiable, customizable intelligence— founded on content-first infrastructure that supports audience reach, engagement, and revenue. By bringing clarity and control to content navigation, Overtone empowers media professionals to align with what drives brand relevance today and brand value tomorrow.

This leads to larger questions about how to govern citizens that on the face of it doesn’t seem to agree on much, and the information environment that could make it easier to have civil discourse about the challenges in public life.

Earlier this month Overtone’s Chief Product Officer sat down with Prof. Chris Summerfield of Oxford to chat about his work on the “Habermas machine” that uses AI to help people, many with conflicting views, to find agreement.

Below you can find updates on the “tones” around the U.S. election as well as updates from Overtone.

Read the post

SWING STATE OUTLETS SHOW DIFFERENT “TONE” TO NATIONAL DISCOURSE

Our last newsletter looked at a tumultuous period in the U.S. election, from a poor debate performance by President Joe Biden to the attempted assassination of former President Donald Trump and Biden’s replacement as nominee by Vice President Kamala Harris.

Since then, Trump won all of the major swing states and is set to return to the White House after a campaign that was notably divisive by the end, though the actual “tone” leading up to the election depended significantly on which sort of media you were reading.

Overtone analysed the tones of all mentions of major candidates in a bundle of 10 liberal-leaning sources, such as The Guardian, Huffpost and MSNBC, as well as a similar bundle of 10 conservative-leaning sources, such as The Daily Mail Online, Fox News and the New York Post. The level of “angry” mentions was as much as 50% higher in the liberal group, with some days leading up to Election Day seeing 6 out of every 10 mentions with that tone, and mentions of “Trump” regularly surpassing that. The conservative side saw a much lower percentage of anger, including comparable anger numbers for mentions of Vice President Kamala Harris (4 out of every 10 angry) seen in liberal sources.

It’s not particularly breaking news that Americans live in different media realities, but what can these results tell us about the way that the election unfolded, which ultimately came down to the results in a handful of states? Overtone also analysed a group of local, rather than national, sources in swing states, such as TribLive in Pittsburgh, Pennsylvania, the Las Vegas Review Journal in Nevada and the Arizona Daily Star.

Subscribe now

These states may have been the focus of intense campaigning and barrage after barrage of political ads, but the actual news they were getting had a different tone than one might expect. In fact, the mentions of major candidates had a significantly less angry tone than was seen in the national liberal outlets and national conservative outlets.

The information audiences of these papers were receiving was far more “informational” rather than putting a particular emotion into the writing. While many voters in these states were receiving information in a national context, they are also citizens of their local communities, and perhaps study of the fact that local media was closer in tone to conservative media would have made results less surprising for some.

One thing to point out in the differences between media outlets is also that they all contain most of the available tones. So just as people contain multitudes, so do media outlets that bring in different audiences to different sorts of articles, from informational explainers to heavy hand-wringing. In order to truly make sense of it you need to look closer (or get an AI to help).

Overtone Updates

Overtone for ads
Overtone is currently integrating with a major platform that will allow advertisers to use our understanding of news to place their ads in the safest, best suited, most resonant spots while supporting journalism. Get in touch to learn more or to be one of our trial clients.

Send us an email

Bluesky

The social network Bluesky has seen a burst in users recently, and offers developers such as those at Overtone options for labelling and sharing articles. Take the poll below to let us know if it would be something we should explore.

Institute for Non-Profit News & Associated Press

Overtone was happy to work with one of our clients, the Institute for Non-Profit News to bring their “On the Ground” initiative to Associated Press members ahead of the U.S. election, using our technology to select and share stories from their network for wider distribution.

Philip at FIBEP
Overtone CEO Philip Allin presented about the use of AI for media monitoring at the Fédération internationale des bureaux d'extraits de presse (FIBPEP) conference this year.

Bauer Media and Bridge AI
Overtone is pleased to be working to test our tone model in the podcast space in collaboration with Bauer Media UK as part of the Digital Catapult programme. Philip will present about the project on Monday in London. Register here.

“You know, a table is only so big, and if you've got to get everyone around the table, then that's going to intrinsically limit the number of people who can be involved in discussion. At this point, a light went on. Maybe you could actually use a language model to solve that problem,” Summerfield said in an interview.

The group created a structure of multiple AI applications working together, one of which generated statements and the other of which predicted how likely the statement was to be accepted by the different human members of the group, allowing it to select the one that would be most acceptable.

Summerfield said he was surprised by the level to which the Habermas Machine incorporated viewpoints that were not the majority in each group. In some ways it goes against a common understanding of language models as repeating patterns seen in their training data because they occur the most often. A piece from the flurry of attention around AI image models last year spotlighted how they create Native Americans or Soviet soldiers with beaming American selfie smiles, presumably because there was large amounts of training data that was similar and despite the fact that Soviet soldiers and Native Americans don’t share the same cultural smile as the 21st century U.S.

“What it seems to do is write statements which are faithful to the majority, but include prominent elements of dissent. So they weave in stuff of people who would otherwise feel kind of disenfranchised. It weaves their views in so that they feel heard. And this is what you hear again and again when you interview people from citizens assemblies, that one of the reasons that people feel positively about the process is not necessarily because the outcome matched their viewpoint. Very often it doesn't, but people feel like they can contribute,” Summerfield said.

The question then, is what exactly would an artificial intelligence solution like the Habermas Machine be scaling if it was used online in a widespread way?

Much of academic media studies and discourse theory is looking at idealized speech situations, where people are logical and measured in their statements, rather than the real rough and tumble tumult of a social network, for example.

Another wrinkle from the study focused on citizens assemblies is that the artificial intelligence machine wasn’t in charge of supplying facts or reading material to the participants. Two people can be influenced against trusting a source or disregard information based on where it is coming from.

It is something that we at Overtone think a lot about when we want our customers to be able to judge a piece of content not only from whether its source is famous or popular or not, but based on what it contains (Is it opinionated? Is it factual? Is it angry? Is it happy?). Perhaps it is easier to get people to agree if they start from the same place, or you at least know which audience they are coming from and the characteristics of the media they are reading.

It is something that Summerfield wants to continue to research in the hopes of finding better methods for making discourse happen.

“Many, many things can be cast as a mechanism design problem. And I do think people are a lot smarter than we give them credit for. People also are a lot more well meaning than we give them credit for. If you find the right modes of expression, then things can be functional,” he said.

“If you take a look, [at the citizen assemblies] one of the things you will see is that the viewpoints that people express are remarkably sensible. They're remarkably articulate. They're remarkably heartfelt. People just come across really well. It's like the opposite of what you would expect if you just scroll through Twitter for a while.”

Measuring how people feel about things online is a tricky business, with much work in AI focusing on what is called sentiment analysis, looking at text to see how someone feels about a certain subject. Classic examples include running sentiment analysis on movie or Amazon reviews to see if someone likes a film or product.

But a lot of sentiment analysis also takes a decidedly unhuman approach to how people feel, and flattens the entire world of people’s emotions to a simple negative to positive scale, like -1 to +1. It also plays a lot on the perceived sentiment of individual words, which leads to silly situations like a leading Big Tech sentiment analysis tool saying that the headline “Drug company slashes cancer risk” is -0.6 for sentiment despite the fact that any human would understand that the headline is saying something positive. 

With that in mind, Overtone has now built a model that looks at tones in a text, focusing at first on categories including, angry, sad, fearful, happy, funny and hopeful. In the drug company example above, our model says that the text has a 99% probability of being hopeful. This move to more human-understandable tones also means that it can be used in places that require more nuance, like news.

A shift in tone

You can use tone analysis on anything from mentions of your CEO to mentions of a meme. In the UK we differentiated between party leaders who had sad, angry and funny mentions (The one with the most sad mentions leading up to the election lost in a landslide). But you can also look at these sorts of tones over time to see the twists and turns of a race as a campaign unfolds.

The graph below looks at thousands of news paragraphs that mention U.S. President Joe Biden, former President Donald Trump, vice presidential candidate J.D. Vance and Vice President Kamala Harris from late June, the week of the Biden-Trump debate.

Our models rated each paragraph as to the percentage probability of the tones within it, and in aggregate the chart tells the story of the events that have taken place, including the attempted assassination of Trump on July 13 (a spike in fear and sadness in mentions) and Biden dropping out and endorsing Harris on July 21 (a surge in hopefulness in mentions). 

This sort of analysis is definitely useful in the media monitoring space, though we are also excited for publishers to begin to use it to tackle news avoidance, where readers regularly and repeatedly say that one of their biggest problems with news is that it is too negative. By understanding tones, newspapers can send the right article to the right reader based on their user needs, and package different sorts of articles together (much like how in the past the physical newspaper had the big scandalous political stories and the comics all in one bundle).

News vs Social

You can also see in the chart above that the predominant “tone” conveys information in a relatively neutral way, which we call informational. That is to be expected given that the paragraphs were from news articles. 

However, not all platforms are like this. We ran our new tone model on posts from social media that we collected, also mentioning the same four candidates, to compare. What you see is one tone dominating above all the rest, and that is anger. 

This finding is particularly interesting from a business perspective, where both news sites and social media platforms are competing for the same pool of ad dollars from companies that want to place their campaigns next to content where readers are paying attention. It appears to confirm that news has a much less angry tone than social media.

The tone of social media also plays on brand safety concerns those advertisers have and continue to have, especially in turbulent political times. They do not want to be associated with content that is offensive or spouting conspiracy theories, and often would want their message next to something more inspiring, like the Olympics. 

Elon Musk has sued a group of advertisers and the Global Alliance for Responsible Media over controversies about brand safety on X (formerly Twitter), though an analysis with Overtone’s brand safety models also shows that social media tends to be a place where the risk is higher. Of the articles and posts mentioning the candidates, the vast majority in all circumstances on both news and social media didn’t mention conspiracies, or discuss/promote violence or offensive speech. However, the percentages in those risky categories were much higher on social media.

News had less than one percent in the High risk category for each of the three risks, though social saw small but significant portions, including more than 7% of all candidate mentions that were High risk for being offensive. If businesses want to avoid being caught up in angry online discussions, and even work towards rewarding a change of tone in online discussions, supporting news would be a good place to start.

If you’re interested in hearing more, please get in touch.

Send Overtone an email

Check out the demo

Insight into leaders

Beyond just using the tool to look at articles, you can also look at the Insights tab to dig further into understanding how coverage of the election is unfolding. On the first graph on the page you can see the different types of articles about the election, as well as those articles over time. The total numbers of articles have gone up and down, though by using the filters to look at the different Overtone “types” you can see, for example, a peak of opinionation about the election on June 27, right after the final head-to-head debate between Prime Minister Rishi Sunak and Labour Party leader Keir Starmer.

You can also see what individual candidates are up to. By filtering down to just articles on Starmer, you can see the potential next leader doing a growing barrage of Interview Stories as the public gets to know him better.

On Sunak’s side there are some interviews, though more noticeable is the growing number of reporting pieces and particularly Investigation pieces, including pieces on the police looking into a scandal that saw the prime minister’s aide investigated for allegedly placing bets on the day of the election before it was called. 

The second tool shows articles that mention conspiracies or conspiratorial language. As we have discussed in previous newsletters, many of these pieces are blocked from receiving advertising based on their keywords, though more nuanced models can differentiate articles that mention conspiracies in the context of news (Low Risk) from articles that blast the “Westminster cabal.” (High Risk)


Introducing Tone

Overtone is pleased to provide here the first look at our new “tone” model that evaluates for emotional tones such as happy, sad, angry, funny, or purely informational. We are still working partners in publishing and other industries to fine-tune the model and the categories it looks at (reach out if you’d like to be involved), though wanted to show how this model can help understand news at scale by looking at how authors are writing about people like candidates.

Below you can see a graph showing more than 27,000 mentions of Sunak, Starmer and the Liberal Democrats’ leader Ed Davey from recent articles. While many of these pieces mention them in purely “informational” ways, there are clear differences in the other emotional tones.

Sunak is the subject of the most sad mentions, with articles like this one from the Financial Times that includes statements like “Only a couple of months later Sunak implicitly recognised the hopelessness of his "change" pitch by bringing back Cameron as his foreign secretary.”

Meanwhile, Starmer is the subject of the highest percentage of angry mentions, with articles both from the right-wing press like this one from The Sun blasting him as a “hard-left socialist” and this one from the left-wing Canary calling him “undoubtedly a safe pair of hands for the ruling class.”

Particularly interesting to see for us as we were developing the model, it was also able to pick up on the fact that Lib Dem leader Davey is often performing stunts and photo ops, and the differences that his coverage receives because of that. Davey, who recently went bungee jumping on the campaign trail, sees a whopping 20% of all mentions of him in a funny or happy way.

Blog post

After attending the Cannes Festival of Creativity last month, Christopher wrote up a blog post of his impressions, including interviews with publishers and agencies thinking about how to move the internet forward. Read it below

Cannes Lions was all about finding out which data will be King of the Jungle.

When heading to the south of France I was thinking about the whole media ecosystem through another animal analogy, the parable of the blind men touching an elephant, where each one of them is touching a different part of the animal and, because of what they are touching, think it is something different.

After attending the conference for the first time last week, I’ve realized that this is a pachyderm with even more parts than I thought possible, with publishers, creators, agencies, tech companies and others all hoping their own view becomes the dominant one.

Subscribe now

The views on what *should* happen are particularly interesting because the industry is in flux, with the impending demise of third-party cookies that is still being talked about years after it popped up on the horizon. Those cookies gave us a workable understanding of what was happening online by gathering a lot of information on people and their personal data, which is definitely an important part of the elephant though not compliant with privacy and not telling the whole story.

Everyone from agencies to publishers is now hoping that their starting point on the elephant is the right one for understanding the whole thing.

Finger”print”

“Quad is a Midwestern company and there is a real sense of transparency, wanting to do the right thing,”  Josh Golden, Chief Marketing Officer of Quad, told me over breakfast at the company’s apartment.

With its headquarters outside of Milwaukee, Wisconsin, Quad is in fact different from a lot of larger agencies because of its background in printing. Huge numbers of publications, from The New Yorker to shopping catalogs, pass through its presses. It’s a volume that Golden, who is also former president of Adweek magazine, tells me reaches 89% of U.S. households. 

Mailbox data may seem very 20th century, though it is also something very concrete. In a world where all parties in the ecosystem want more clarity about why their money is being spent where, having a (excuse the pun) paper trail of concrete, human-understandable data can be very valuable.

“The household mailbox is the most resilient piece of information you have. Now we’re reverse engineering it for the digital side,” Golden’s colleague Howard Diamond, Chief Growth Officer of Quad’s agency Rise, said. 

The idea is that by starting with a set of data that is solid and offline, you can get a better understanding of online behavior and business. Quad’s emphasis on “household” data for advertising strikes me as interesting because in some ways it seems similar to a big push from publishers on their “first-party” data, like email addresses or story preference that they get from users directly through their trusted relationship of outlet and reader.

But then there are what seems like a million companies out there in the “AdTech” space saying that it is actually their data, or their way of combining lots of data points together in some novel way, that is the answer to all your problems. There are trends like viewability of ads or biometric data or attention data. You can stuff a thousand pieces of data in a box much like financial traders stuffing subprime mortgages into a product and then having it rated AAA. It’s unclear which data, on which channels, of which type, will actually prove to be valuable.

Content cohorts

I met Renn Turiano, Senior Vice President and Head of Product at Gannett, when he was on a panel at the Yahoo! beach on making sense of a “fractured media ecosystem.”

“There are a lot of replacement strategies [for cookies] that are in the works right now, some sponsored by Google and from others as well. We think that as an organization, of course we'll lean into all of those, and we'll follow whatever the market follows in terms of what the ad community chooses to embrace,” he told me in a follow-up interview. 

“But we're not going to leave it there. And we're not going to wait for all that to happen to us. We're pushing forward with strategies that we hope will develop more and more direct relationships with consumers.”

That direct, trusted relationship with the audience is unique, and even more so because publishers also create and control the articles being read, which is what many people go online for in the first place.

“If you understood all the attributes of a piece of content, as well as you hoped to understand the attributes of the individual consuming it, then you'd really be in business. So you could definitely create as many nuanced and certain subset cohorts of content that you like,” Turiano said.

Content itself is an unavoidable, essential part of the elephant. The question then is what else you can do with data on content, and how it can be extrapolated to get you to understand other parts of the elephant.

“At Amazon they say, ‘We can get gender in three clicks’ from an anonymous user. And that's because of everything that they know about the referrer, when they come into the domain, and everything they know about each thing that they chose to click on,” said Turiano, who was formerly a creative director for the online marketplace. 

Lions and elephants

There are many places to start from, but my time at the festival left me convinced that not all starting points on the elephant are created equal, and that ideally, like the real lions the Cannes awards are named after, you respect the full elephant rather than trying to take a piece out of it. 

There is data about platforms and all sorts of data about people, but to paraphrase Walt Whitman, people contain multitudes. You can be a mom in the morning, a c-suite executive in the afternoon and an indie rock fan in the evening. Content and data about content is the thing that can help you see the whole picture, with all its nuances. 

If you’re interested in understanding the entire elephant, rather than just carving out a niche in it, content is the only place to start.

The more I dig into the media and tech ecosystem, the more the parable of the blind men and the elephant comes to mind. Not seeing the full animal, one blind man touches the trunk and thinks it is a snake. Another touches the tail and thinks it is a rope. Still another touches the side and thinks it is a wall.

Early summer is the height of the event season, and we can learn a lot about the beast that publishers, platforms and advertisers each think we are dealing with by seeing who is mingling with who.

At the time of Overtone’s last newsletter I was off to Italy for the International Journalism Festival in Perugia. Journalists, editors and decision makers from around the world discussed press freedom, funding, reaching audiences and artificial intelligence. 

Perugia skews heavily towards editorial people. Overtone sometimes calls them ‘red’ people in the pyramid below; ‘blue’ people are mostly on the business side. Blue and red people often have no interactions – even when they work at the same publisher. They also have separate events, with one of the big blue ones being the Cannes Lions Festival of Creativity. Overtone will be there next week – drop us a line to say hi!

Maybe because Overtone is deeply concerned with seeing the whole elephant, I am happy to be going to both conferences this year.

But it is a very small number of people who do that, and that seems odd. Blue people are crying out for a better understanding of editorial workflows. Meanwhile, red people are in dire need of sustainable revenue streams. So Overtone is doing its best to learn from both, and to share with both.

The bright beautiful cookieless future

The rise of generative AI has excited and scared the editorial side of publishing, though it also has come at a time when things are in flux for the blue side and its advertising partners. Right now we are seeing the slowly unfolding demise of cookies that suck up people’s personal data, which means that advertisers are on the hunt for places where their ad campaigns can be effective in a more privacy-focused way.

Some major players in the advertising space such as The Trade Desk are putting forward the “premium internet” that takes ad slots from a top 100 group of publishers and sells ads there based on other forms of data, including “first-party data” from the publishers such as the email address someone registers with. 

There is also “contextual” data about where the ad is being placed, such as Overtone data about the content itself. I expect to have both of these topics to be big at Cannes and am particularly interested in where these conversations touch on AI. Before when all you had was a hammer (clicks, shares, keywords) everything was a nail (a way to count those metrics, like in cookies), but now there is an entirely new box of tools.

Brand safety and keywords

One of the big trends for the advertising side that editorial people should also be aware of is that not insignificant percentages of articles that could have advertising are blocked from receiving it because of concerns about safety for the brand’s reputation.

Brand safety has been on the rise as big companies and their agencies want to avoid things like promoting and funding conspiracy theories, or putting their campaigns in a place that would create a bad impression. Think of an airline that doesn’t want to tell you about its new deal on summer flights next to an article about a plane crash.

It’s a very reasonable idea, though has run into some issues in practice, with keywords added to blocklists that end up cutting off huge numbers of articles from being funded. If you blocked the keyword “shooting” for example, you would block some crime articles, but you could also end up blocking huge amounts of articles about football and the upcoming Euro 2024 tournament.

Even when the blocking systems work as intended, the fact that they use keywords, and keywords are a very blunt instrument, means that the baby can be thrown out with the bathwater.

The three articles below all come from a reputable publisher, CNN, and all three of them were blocked because of “Conspiracy Theories.” But when you look at the articles themselves you can see that they are not a monolith and there is additional nuance needed. This article about Taylor Swift conspiracy theories around the Super Bowl quotes extensively from conspiracy theorists talking about the pop star. Overtone’s systems, based on the Global Alliance for Responsible Media (GARM) standard, say that it is high risk because even if CNN is reputable, actual conspiracy theories are being shared and advertisers may not want to be a part of that.

The other two however, do not actively share conspiracy theories. CNN’s article about Rudy Giuliani’s radio show being canceled makes mention of conspiracy theories that led to court cases against the former mayor, but does not delve into them. The article is essentially about the radio show and Overtone’s model rates it “Low Risk” according to the GARM standard.

The third article shows just how far keyword blocking can go, to the point of being a little silly. Ryan Gosling appeared on Saturday Night Live in a sketch about alien abductions, but the article was blocked on the grounds of “conspiracy theories,” presumably because of the aliens. Overtone would rate it No Risk, and most advertisers would probably be perfectly fine having their ad on that article.

Questions are being raised about how to make this whole process smarter, and Cannes will be the place to have them. If you are there, please reach out to have a chat!

Get in touch!

Announcements

Overtone has redone our website! Head over to https://overtone.ai and let us know what you think of our new look (and new logo!)

By Christopher Brennan

Imagine your job is to fill up the advertising spots on a subway train. You are proud of the service that you offer, getting people from place A to place B, and you want to find $100 from advertisers who can help support you, pay the conductor, and keep the train rolling.

The big businesses in town are rubber companies.

The local agencies come to you and say they have posters they want to put up for new tires. But they have a hard time proving that their advertising is working, and they have a lot of other places they could put it, so they don’t want to pay much for each ad.

You need a way of making sure you can get your $100 so you start increasing the number of ads in each subway car, from just a couple to eventually covering all of the seats and handles. At one point you start searching riders’ pockets to see if you can prove they own a car. It’s still not enough.

Eventually, you decide to build more subway cars and attach them to the back of the train, filling them with tire advertisements that are not working in the hope that quantity will fix a problem of quality.

This sounds silly. Why advertise tires to commuters who are actually just tired? Why doesn’t the subway guy advertise products that might land better with the audience, say inflatable rubber mattresses? But scenarios like this are still playing out in digital publishing. There is still a gap between the audience that advertisers want to reach and the audience they do reach.

Forbes scandal

Which brings us to Forbes. With the industry at large facing declining ad revenue, Forbes began to sell ads on a secret, second site called www3.forbes.com. Major companies like Procter & Gamble, Disney and McDonald’s all paid for ads on what they thought was the real Forbes, but what was in fact the second site with a different audience, according to reporting from Adalytics and the Wall Street Journal. Previous research has shown that showing ads on this type of clickbait-y ‘Made for Advertising’ sites is essentially worthless in helping those companies hit their goals.

Sketchy Made for Advertising stuff is not new (we wrote about it here) though the fact that Forbes was accused of doing it has had a huge impact in the ad world. It was shocking only to those who haven’t seen the struggles of digital publishers to connect with an audience.

The surprise is not the alleged Forbes strategy, the surprise is that we haven’t seen more people trying it in desperation, focusing on maximizing one metric, revenue, at all costs.

So what happened to Forbes in the digital age?

Diving into Forbes’ output tells us a lot about the landscape they have operated in over the past 15 years. Overtone ran 1,000 articles on Forbes.com from around 15 years ago, before major management changes, and more recently. First, let’s look at a breakdown of the story types from these two distinct time periods. Some changes in content type are to be expected – online publishing has changed a lot over the past 15 years – but others point to red flags.

Take the next chart, below. The overall number of news stories in the samples hasn’t dramatically changed, so looking at blunt article tags like ‘news’ or ‘business’ won’t sound any alarms. But by analyzing each article’s content we noticed a huge swing from deep-dive news reports and investigations towards more general, resource-light, reporting.

It’s similar with opinion. An Overtone content analysis shows a dramatic shift from deep think pieces to quick-hit hot takes.

That change in article types and decisions with www3.forbes.com are intertwined. Highly clickable articles were distributed via recommenders on other people’s sites. This was done to make as much money as possible from any in-bound traffic. Some of those pieces we traced are, predictably, not very substantive.

It’s easy to see how a publisher can tip into a downward trajectory, reducing quality to produce more content to chase ad revenue based on clicks. But then ad revenue slumps because the quality is low, so the publisher needs to produce more content. Limited resources mean that content is more and more like filler, so audiences disengage. And that’s how a publisher gets caught in a death spiral.

This cycle would be broken if publishers could offer some of the crucial insight they have into their own audiences for use in the systems that run advertising. But that cannot happen until publishers and advertisers alike have more substantive analysis of content.

A way forward

Into this world of competing for clicks has come generative AI, where it is easier than ever to flood the zone with content designed to game these advertising systems without really caring about the reader. We are now in a world where AI entrepreneurs can openly boast about stealing web traffic through creating low quality AI content at scale.

Yes, that’s a threat to publishers but it’s also an opportunity to demonstrate the value of well-researched, thoughtful work.

Artificial intelligence presents an opportunity to fight fire with fire. By having AI models fine-tuned to “understand” nuances in articles, it plays a role in rooting out bad practices such as Made for Advertising, or conspiracy theory websites that no reputable company really wants to support with their ad dollars.

It also creates an opportunity for publishers to really show off what they have, which is insight into their articles and their audiences. Whereas before technology could only assess articles at the click or keyword level, more advanced AI models can now unpack what an article is trying to do and the experience someone is having while reading it.

This is something advertisers want that has been missing online. To return to the subway analogy, when an ad is shown on the subway, the person making that ad knows exactly what experience the passenger is viewing it is going through, like a morning commuter looking up at a banner for an energy drink that asks “Tired?”. That level of caring about the content is almost completely absent in the online ad ecosystem, where keywords provide very little insight into what someone is reading and the main way of serving ads is by sucking up as much personal data as possible and tracking someone around the internet.

With language models a new way, more like the subway, is possible, and it plays right into the knowledge that publishers, who are sitting on a gold mine of insight, already have. All of this while having a better relationship with readers and letting their personal data remain private.

Tone

Some outlets are already looking into this. The New York Times is working on using AI for ad placements and is currently in testing, Axios reported. In Spain Henneo is also working on personalised advertising with AI.

For everyone else, Overtone is pleased to announce that we were awarded an Innovate UK Smart Grant to fund creation of a “tone” model that can look into types of articles like tragic, sensational, inspiring or humorous. Not only can this help publishers understand how to combat problems like news avoidance (more on that soon) but it can also refocus the ad side of the business on something more tangible and human.

We see this as an important next step in helping publishers unlock the value they have that has been lost in the world of clicks and keywords. Overtone is putting together a working group of publishers and other interested parties, starting with an event in London this spring and continuing for 2024. Click the link below and send us an email if you’d like to be a part.

Send Overtone an email to get involved

Interview

Alessandro De Zanche is an expert in data, advertising, and audience strategies. He worked at News Corp and Yahoo!, and now consults for different clients after seeing the industry from all sides. His Not Just Adz newsletter is a must read. You can check out our interview with Alessandro here.

Announcements and Events

Christopher is currently in Perugia, Italy for the International Journalism Festival, with support from VRT Sandbox and the Seeds & Growth programme. Reach out to him if you are there!

Newsletter assembled by Christopher Brennan and Kevin Donnellan

By Christopher Brennan

Over the past several months you may have noticed a new sort of piece on your favourite tech or media website: the AI sludge story.

They delve into weird corners of the internet, where shady characters sometimes make hundreds of thousands of dollars through quirks in the way that articles on the internet are distributed and how those articles receive attention – and money.

One of the first examples was from late last year when Jake Ward, an Search Engine Optimization specialist, bragged about pulling off an “SEO Heist” where he used generative AI to programmatically create thousands of articles based off of a competitors headlines. He would post the articles on his client’s site and then watch as it sucked up search engine traffic and brought thousands and thousands of eyeballs.

Another mentioned AI-generated articles appearing in ‘news’ search results or AI-generated “obituary spam” about individuals who were in fact, not dead.

Perhaps one of the most intriguing was “Confessions of an AI Clickbait Kingpin” from WIRED about a Serbian DJ and entrepreneur named Vujo who buys up old websites and is now using AI-generated articles to suck up SEO clicks from search engines, amassing what he says is hundreds of thousands of dollars for some sites. Christopher, Overtone’s CPO, sat down with WIRED reporter Kate Knibbs this week, and you can read the interview below. 

Are we entering the age of AI Clickbait Kingpins?

Read the interview

But these stories also point to one of the reasons that Overtone was founded, to redefine how value is measured online beyond just clicks and shares. Generative AI has made this harder (but also more worthwhile), and we risk losing some of what is human about being online if we can’t catch up with it.

It also pinpoints how using fine-tuned AI language models like Overtone’s can be beneficial for dealing with the scale of some of these issues.

On the Wired piece, for example, one of the sites that Vujo bought was The Frisky, which used to be a popular women’s entertainment site in the early 2010s, the golden age of witty, sometimes snarky, opinionated blogs like Gawker. We ran hundreds of articles from both the old Frisky and the new Frisky, post-Vujo acquisition, and you can see the difference. The old site had typical pieces like “This Thanksgiving, Eat Whatever You Goddamn Want” by Lauren Duca, a real person writing a takedown about diet advice in women’s magazines. If not Pulitzer-winning work, it was more considered than the new site which publishes fluffy low-depth “hot take” pieces like “Why Escorts Are the Perfect Companions for a Night Out” or “Dental implants – What you need to know before getting one.” As Knibbs mentions in her interview, it seems like the pieces are “not really intended to be read at all” but just suck up search traffic. If you wanted to know when a site became something like this, you would either need to monitor it manually or use something like Overtone AI signals to understand when the site became something unrecognisable that you may want to stay away from.

Or take the SEO Heist Guy, who generated thousands of articles for the website of a financial planning tool. More than 90% of these generated articles were low-depth filler, which would be easy for an AI system like Overtone to detect and have advertisers or monitors pass over. Even if it is a corporate blog, it is particularly lacking in substance with pieces rambling explanations of Excel functions or “Agile CRM vs Zendesk Sell”. By comparison, take the blog of a company like Rakuten that publishes something like “How AI is changing creativity: A physicist weighs in”. While it is definitely in the business of promoting Rakuten, it is offering something that is meant to be read by a human.

That question of being “meant for humans” is ultimately one of the biggest and thorniest issues to tackle as the internet fills up with the flood of content. It points to how the current systems of moving articles online are focused almost entirely on impressions, clicks and shares. AI for reading can create a different way of looking at things.

Overtone Press

Overtone’s CEO Philip Allin sat down with the newsletter Quo Vadis from Tom Triscari to talk about the problems we are tackling including “Made for Advertising” sites.

Overtone Events

Philip will be at the Mobile World Congress next week in Barcelona, as well as at the 4 Years From Now side event.

In the world of publishing Reagan will be at the International Symposium on Online Journalism in Austin and Overtone will be participating in the MediaTech Festival in Odense, Denmark.

We will also be presenting at Tech Day for the media intelligence association FIBEP in April in London.