ChatGPT now shows ads in Europe. What changes, and how to avoid them
ChatGPT ads went live across 31 European markets on 24 August 2026. Who sees them, what data they use, and how to get an ad-free answer.
A sponsored card, on a Tuesday evening, in Ghent
On 24 August 2026, ChatGPT began carrying advertising across 31 European markets: all 27 EU member states plus Iceland, Liechtenstein, Norway and Switzerland. Nothing inside the answers changed. Something around them did.
Ask the free version which laptop would suit a design student on a modest budget, on a Tuesday evening in Ghent, and the reply arrives the way it always has, three or four options with reasons attached. Underneath it now sits a small card, carrying a product image, a line of copy, a link to a retailer, and a label that reads Sponsored. That thin strip of screen is the whole story.
OpenAI had announced the expansion the week before, on 19 August, and the launch was covered by EU Perspectives, Digiday and FashionNetwork among others. Germany, France, Spain, Italy, Poland, Sweden, Denmark, Ireland, the Netherlands and Austria were all named. Europe got the full product, with a different set of rules attached to the data behind it.
A week later, on 31 August, OpenAI said the ads business had reached a billion dollars in annualised revenue roughly 200 days after the first pilot. Forbes, Adweek and Quartz all carried the figure. OpenAI's chief marketing officer, Colin Fleming, added the number that explains why: ChatGPT has more than a billion weekly users, and around a fifth of them show what he called commercial intent. One in five people asking a machine a question is, somewhere in that question, thinking about buying something.
Who actually sees them
This is where most of the panic online is misplaced. Ads reach logged-in Free and Go accounts. Plus, Pro, Business, Enterprise and Education remain ad-free, and OpenAI's stated principles exclude accounts identified as under 18, which now sits alongside the ChatGPT for Teens mode launched on 18 August.
One detail carries more weight than it looks. Go, the cheap paid tier at 7.99 euro a month in Europe, still shows ads. Paying something does not buy you out. Paying at the Plus level does. If you have been weighing up Go on price alone, fold that into the decision, and we went through the rest of that tier in our look at whether ChatGPT Go is worth it.
The format itself is restrained, at least for now. The core unit is a sponsored card rendered below the model's answer, visually separated from it, with a title, a short description, an image and a destination link. Retail advertisers can run product feeds. A multi-product carousel entered testing during August. OpenAI's position is that ads do not influence the answer above them and that advertisers receive no conversation data. That position is the company's own account of its own system, and until later this year there is no way for anyone outside it to check.
The Peloton story you have probably read is from the wrong year
Almost every article about this gets one thing wrong, and it repays slowing down for.
Search for ChatGPT ads and you will find dozens of posts telling you that OpenAI showed ads for Peloton and Target inside unrelated conversations, and that even people paying two hundred dollars a month for Pro got them. The story is real. The date attached to it usually is not.
That incident happened on 7 December 2025, and it involved a feature that surfaced unsolicited app suggestions from partners inside conversations where nobody had asked for them. It had nothing to do with the paid advertising programme, which did not exist yet. Users read them as ads, because they looked like ads. OpenAI switched the feature off within days. Nick Turley said at the time that there were no live ad tests running. OpenAI's chief research officer, Mark Chen, was blunter.
"I agree that anything that feels like an ad needs to be handled with care, and we fell short."
Paid advertising was announced separately on 16 January 2026, and the trade press timeline puts the US pilot live on 9 February, a date reported by aggregators rather than one we could confirm in a first-party post. Set the two events beside each other and there are nine weeks and a formal product launch between them. So when a blog recycles the Peloton screenshots as evidence that ChatGPT is already pushing ads at Pro subscribers in 2026, it is describing a feature that was withdrawn before the ads programme existed. The correction matters more than the outrage, because the real programme has held its tier boundaries so far, and any honest criticism of it has to start from what it actually does.
The data behind a European ad
At launch, European ads are non-personalised. According to Trending Topics, which reported the detail ahead of the rollout, the targeting signals are the topic of the conversation you are currently in, your approximate location, your device type, the time of day and your language. Past chats and stored memories are explicitly excluded.
So in the Ghent example, the system knows someone in roughly that area, on a phone, in the evening, writing in Dutch, is talking about laptops. It does not, at launch, know that you asked about a divorce lawyer in March or that your memory file says you have two kids. That line is drawn where it is because of European law rather than product philosophy.
Personalisation is planned for a second phase and will require you to opt in. That is the moment to pay attention. The gap between an ad matched to a sentence you just typed and an ad matched to everything you have ever typed into an assistant is enormous, and the second one arrives as a checkbox rather than an announcement.
The accountability chain is unusually clear. The data controller is OpenAI Ireland Limited, based in Dublin, which makes the Irish Data Protection Commission the lead supervisory authority for GDPR questions about all of this. There is also an unresolved legal argument, flagged in the same reporting, about whether a free tier with ads sitting next to a paid tier without them satisfies European guidance on consent-or-pay models. That question has been fought over with social networks for years and has now arrived at AI assistants.
Then, on 31 August, the European Commission designated ChatGPT a Very Large Online Search Engine under the Digital Services Act, the first AI chatbot ever designated. Among the obligations that begin four months later is a public advertising repository. From late this year, in principle, anyone will be able to look up which ads ran inside ChatGPT and who paid for them. An assistant that answers questions and sells ads against the answers is now required to show its working, which is where that unverifiable claim about ads not influencing answers finally gets an outside check.
One thing to keep separate: none of this is the EU AI Act. The AI Act's transparency rules started applying on 2 August 2026 and cover disclosure that you are talking to a machine, along with the labelling of AI-generated content. They place no obligation at all on the sponsored card. Plenty of coverage blurs the two.
Getting an ad-free answer
The practical options are narrow and better stated plainly. Move to a ChatGPT tier that excludes ads, which means Plus or above rather than Go. Go into settings and decline ad personalisation, which does not remove the sponsored cards but does limit what feeds them, and assume that toggle becomes more important, not less, once the opt-in phase arrives. Stay alert to the case where sponsorship and advice sit closest together, which is any question about what to buy, book or subscribe to. Or use an assistant that does not run ads at all.
We fall in that last group, and it is the only claim we will make here: MultiChats has no ads on any plan, free included, and no plan to add them. The pricing is on the homepage if you want to check that against what you are paying now.
The judgement
OpenAI has built the least objectionable version of a thing many people did not want. The label is clear, the card sits below the answer rather than inside it, the paid tiers were not quietly reclassified, and the European launch really did ship without touching your chat history. Give it that much.
But a billion dollars in 200 days is the loudest signal in this entire story. Advertising on that trajectory does not stay in a card below the fold. It grows by becoming more personal and harder to see, because that is the direction the revenue curve rewards. The opt-in personalisation phase is the first test of that, and the DSA ad repository will be the first place anyone outside OpenAI can check the answer.
Until then, judge it on the sponsored card rather than on the December 2025 screenshots. And when you ask an assistant what to buy, remember that on some tiers, in 31 European countries, that question is now worth money to somebody.