ChatGPT is opening its doors to advertisers in Europe, AI agents are beginning to buy media from one another, and creators are being trained to write not only for people but for machines. Welcome to the day advertising stopped chasing the gaze and began chasing the decision.
Yesterday, 24 August, somewhere between the first coffee, the emails already piling up like unpaid bills and the familiar Greek ritual of ‘let’s see what burned while we were asleep’, something changed in advertising that most people probably did not notice. No servers went down. No new social network full of dancing teenagers appeared. No new ‘guru’ arrived to explain why the funnel had died — for the eleventh time. OpenAI simply opened ChatGPT Ads in 31 European markets, including Germany, France, Spain and Italy, initially through its Ads Solutions team, agencies and technology partners. Ads concern Free and Go users, while Plus, Pro and Enterprise remain ad-free. Yes, it may sound like another expansion of an advertising product, but it is not. Advertising does not appear here while you kill five minutes watching cats, fit men and politicians who have suddenly discovered TikTok. It appears while you ask, search, compare and have already taken your wallet out of the pocket of your mind, trying to decide where to put it.
That is the crack through which the rest of the story enters. Until yesterday, much of digital advertising lived by chasing attention. Someone scrolled while armies of pixels, cookies, lookalikes, retargeting engines and media buyers ran behind them like the men in Monastiraki who follow you down three streets because you glanced at a pair of shoes for half a second. ChatGPT changes the geometry. OpenAI itself describes the product as a way for advertisers to reach people while they ‘explore options, compare and make decisions’. Not exposure. Intent. And intent here is not the word you typed into Google at 11:43 at night. It is an entire conversation: ‘I need a family car, I have two children, I drive 20,000 kilometres a year, I do not want diesel, I have this much to spend — what should I buy?’ If old search advertising heard a word through the door, conversational advertising may already be sitting at the table while you describe the problem.
While we are still debating whether this is ‘a new placement’, OpenAI is building something much more ordinary and much more serious behind it: a real advertising machine. Digiday reported that the company is looking for a head of enterprise ads marketing to build content, thought leadership, events and demand programmes for its enterprise advertising pipeline. The job description itself says the goal is to position OpenAI as a ‘premium advertising platform’ for marketers worldwide. Meanwhile, Ads Manager is gaining custom audiences, budget alerts, account-health indicators, reporting tools and automated suggestions for headlines and ad copy. These are not the toys of a startup trying a little advertising to see how it goes. They are the plumbing of a platform preparing to carry serious media budgets. OpenAI has already added conversion optimisation, geo-targeting, custom audiences, Pixel, Conversions API and measurement integrations. In other words, beneath the clean white interface where someone asks how to paint their living room, an engine room is being built that smells increasingly of Google and Meta.
At the same time, something even stranger is happening: in this new marketplace, people may soon not need to speak to one another at all. Butler/Till and iHeartMedia completed a four-week experiment this summer buying streaming-audio and podcast inventory in which two AI agents — one for the buyer and one for the publisher — communicated through MCP. A person gave the brief, then the buying agent could move to the publisher agent to execute the media buys. Total spend: under $10,000. Small, almost pocket money by global-market standards. Yet the point is not the amount; the transaction happened. Brief → agent → publisher agent → activation. According to the published data, the experiment achieved a CPM 42% lower than the advertiser’s benchmark for direct purchases, while 48% of podcast impressions went to premium non-skippable mid-roll positions, compared with 33% in the traditional comparison plan. One test does not write history, but sometimes it shows you where the road is turning.
This is where things become interesting for agencies. For decades, entire ecosystems have been built around people taking a brief, opening spreadsheets, talking to publishers, negotiating, trafficking, watching pacing, correcting budgets, writing reports, rewriting reports because the client wanted the green a little greener, and finally gathering on Teams to explain why CPM rose 6%. Now a model appears in which part of this chain can become machine-to-machine. Not tomorrow morning, and not without human oversight. Anyone claiming that media planners are finished in two years is probably selling a keynote or a consulting package. But the operational part of the work is changing in front of our eyes. Butler/Till has already run agentic pilots in other channels, while iHeartMedia says it intends to extend the model into broadcast radio. The question is not whether the agent will replace the media buyer. It is which media buyer will learn to manage agents before finding themselves performing the work those agents do faster.
Then comes the dirtiest problem of all: who gets credit when nobody saw anything? The IAB has already placed two issues in its 2026 initiatives that three years ago would have sounded like a conversation after the third whisky at a technology conference: a blueprint for attribution in the age of agentic commerce and a framework around ‘advertising to AI agents’. The IAB now describes purchases influenced by conversational interfaces, recommendation engines or agents without the traditional trail of an impression and a click, examining categories such as agent-initiated and agent-recommended conversions. That is an earthquake for an industry trained to fight over the last click as if it were a family plot of land. What happens when you tell your personal agent, ‘find me the best hotel in Rome under €250, with parking and good reviews’, and it searches, evaluates, rejects, compares and gives you one choice? Who influenced the purchase? Which message counted? Which content entered the data it consulted? Which media budget won, and who paid without ever appearing before your eyes?
Creators return to the stage here, except this time they are not holding a ring light. They are holding value for machines — or at least trying to. Brands and agencies are already adding a new requirement to creator briefs: AI visibility. Digiday has recorded cases in which agencies examine which creator content appears or is mentioned in LLM answers, while companies begin testing creators as part of their authority and discoverability strategy in AI-search environments. Creators themselves are building AEO sites, adding structured data, transcribing content and checking whether their references appear in AI responses. Suddenly PR, SEO, influencer marketing, branded content and editorial authority are melting into the same pot. For years we asked: ‘How many views did it get?’ Then: ‘How many clicks?’ Then: ‘What ROAS?’ Now a much stranger question is added: ‘When somebody asks a machine which company to trust, will it remember us?’
Nobody should rush to buy a ‘GEO expert’ sign from the internet and hang it outside the office. The market is still full of noise, uncertain measurement and promises that smell as much like 2026 as ‘number one on Google in 30 days’ smelled like 2009. Publishers and agencies admit there is no single standard for measuring AI visibility yet, and different tools produce different results. That does not cancel the shift; it confirms it. Whenever a market begins arguing over how to measure something, it usually means it has already decided that thing has value.
And so, within a few days, the pieces fall onto the pavement and form an image that is hard to wash away. ChatGPT is acquiring advertising inventory precisely at the moment of intent. OpenAI is building the enterprise media stack to sell that intent. AI agents are beginning to buy inventory from other AI agents. The IAB is trying to decide who receives attribution when the human leaves the middle of the journey. Creators, publishers, PR people and SEO teams are starting to fight over who will exist in the memory of the systems that will make the next recommendations.
So this is not merely a new advertising channel. It is a change of pavement. Old advertising stood outside the shop shouting at passers-by. The next kind does not even need to shout. It will be inside the conversation before you reach the shop, inside the data the agent reads, inside the content the machine considers trustworthy and inside the transaction that two systems complete while we leave the phone on the table and drink our coffee.
The real media war ahead will not be fought only over who wins our eyes. It will be fought over who wins the machine that decides what those eyes deserve to see.