ChatGPT Ads Arrive in Europe, Starting With the Free Tier

Open ChatGPT in Europe on Monday and, if you’re on the free plan, you may notice something new sitting near the answers: an advertisement. As of 24 August, OpenAI is switching on ChatGPT Ads across 31 European markets — the company’s biggest advertising expansion yet — six months after it began quietly testing the idea in the United States. Free and Go users are in; anyone paying for Plus, Pro or Enterprise is not.

Answer first, because the shape of this matters more than the noise. This is the moment the “free” AI assistant becomes an ad-supported one for most of Europe, and the detail that will catch people out is the opt-out. You can turn off ad personalisation — which changes which ads you see — but you cannot turn off the ads. The only switch that removes them is a paid subscription. To OpenAI’s credit, it has been unusually careful about the privacy design, and we’ll give it that in full. But “free, with ads you can only escape by paying” is a different deal from the one most people signed up to, and it deserves to be read slowly.

We like a lot of what OpenAI ships, and an ad-funded free tier is a defensible way to keep a genuinely expensive product free. The questions worth asking are narrower: how honest is the opt-out, how private is the targeting really, and what happens the day an advertiser’s interests and your question quietly diverge.

What actually switches on this week

The facts are not in dispute, because OpenAI announced them itself. In a post dated 18 August, the company wrote: “Six months after we began testing ads in the U.S., we’re bringing ChatGPT Ads to 31 European markets,” naming Germany, France, Spain, Italy, Sweden, Norway, Denmark, the Netherlands and Austria among them, with reporting adding Poland and the Benelux countries. Crucially: “As in our existing markets, ads will be shown only to users on the Free and Go plans. Plus, Pro, and Enterprise subscriptions will remain ad-free.”

The pilot began in February in the US; OpenAI says it has since expanded to eight further markets, so the European rollout takes the total to around 40. This is not an experiment any more — it is a platform. In the same announcement OpenAI describes the machinery it has built: bidding “beyond CPM and CPC” to support conversion optimisation, plus “geo-targeting and custom audiences” and measurement “through the OpenAI Pixel, Conversions API, and third-party measurement integrations.” If those phrases sound familiar, it’s because they are the standard furniture of the online-advertising industry, now being assembled inside a chatbot.

The opt-out that isn’t one

Here is the line most coverage skated over, in OpenAI’s own words: “People also have control over ad personalization, with ad-free paid plans available for those who prefer not to see ads.” Read that twice. Personalisation control changes the relevance of the ads. The absence of ads is a separate thing, available only on the paid plans. So the honest translation of “you’re in control” is: you may choose between ads tailored to you and ads not tailored to you, and if you want no ads at all, that will be €23 a month for Plus.

None of this is hidden — it’s stated plainly — but it’s the kind of plainly-stated thing that a hurried user reads as “I can switch ads off,” when what they can switch off is the tracking that makes ads relevant. It is a familiar move: give the user a real control that sits next to the control they actually wanted, and let the resemblance do the work. If you object to ads on privacy grounds, the personalisation toggle helps; if you object to ads on principle, only the subscription does.

The privacy stance, taken seriously

Now the genuinely good part, because there is one and it would be unfair to skip it. OpenAI has not built the surveillance-advertising model that hollowed out the rest of the web. Its VP of Ads, Dave Dugan, has said plainly that advertisers “will not have access to users’ chat histories” and that “information from conversations would not be shared with advertisers.” The company’s announcement adds that it keeps “conversations private from advertisers and never sell[s] customer data,” that ads are “always clearly labeled and separate from ChatGPT’s answers,” and that “advertising does not influence the answers ChatGPT provides.”

Compared with the ordinary bargain of the ad-funded internet — where your behaviour is the product and the line between content and advertising is deliberately smudged — that is a materially better starting point, and we’ll say so. A labelled, separated ad that doesn’t hand your transcript to a media buyer is about as good as advertising gets. If OpenAI holds that line, this could be one of the less objectionable ad businesses in tech.

The caveat is the word “if”, and the stack described above. “Ad personalisation,” conversion optimisation and custom audiences all require the system to know something about you to work; the promise is that advertisers never touch it, not that no profile exists. That’s a real distinction and a reassuring one — but it depends entirely on OpenAI’s internal discipline holding as the commercial pressure to make the ads perform grows. The history of advertising businesses is a history of that discipline eroding one quarter at a time, which is exactly why every AI is so hungry for your data in the first place.

Why a sponsored answer is harder to spot than a sponsored link

The deeper worry isn’t the banner-style ad OpenAI is shipping now; it’s the medium. A chatbot answers in a single, confident, conversational voice, and that voice is the most persuasive real estate on the internet. A sponsored link announces itself as an ad. A recommendation that arrives “in the same conversational tone” as the surrounding answer — the phrase researchers quoted to Euronews used — is harder to recognise as persuasion. OpenAI says today that advertising “does not influence the answers”; the line to watch, for years, is whether that wall between the answer and the ad stays load-bearing.

We’ve made this argument about search and it applies double here. The reason answer engines keep drifting toward the same cliff as search is that once you need to monetise a single answer, the pressure to shape it — to favour a partner, to slip a sponsored recommendation into what reads as neutral synthesis — is precisely the pressure that degraded the thing they set out to replace. OpenAI is starting from a better place than most. But it is starting down the same road.

Europe is the hardest place to try this

There is a reason a European rollout is more fraught than an American one. The EU AI Act specifically prohibits AI systems that use manipulative or deceptive techniques to distort behaviour, or that exploit people’s vulnerabilities — and a persuasive conversational agent that also carries ads is, at minimum, a system regulators will look at closely. Separately, the European Commission has been weighing whether ChatGPT’s search function should count as a “very large online search engine” under the Digital Services Act, a designation that would bring audit and ad-transparency obligations. Launching an ad business into that environment is a statement of confidence; it is also an invitation to scrutiny that a purely US product never faced.

To be fair, this cuts both ways: Europe’s rules are exactly why OpenAI’s careful, no-selling-data, clearly-labelled design is the sensible way to enter, and the company seems to know it. The regime that makes the launch risky is also the regime most likely to keep it honest.

The steel-man: someone has to pay for the free lunch

The strongest case for OpenAI is simple and true. Running ChatGPT for hundreds of millions of free users costs a fortune, and the alternatives to advertising are worse for most people: charge everyone, or degrade the free tier until it’s useless. Ads that fund genuine free access — while leaving paid tiers clean — are a reasonable way to keep a powerful tool in the hands of people who won’t or can’t pay. OpenAI’s framing, that ads “support free and low-cost access,” is not spin; it’s the actual economic logic, and it’s the same logic that kept web search free for two decades.

Concede all of it, and the objection just gets sharper. It isn’t “OpenAI shouldn’t run ads.” It’s that the free user pays in two coins at once — attention and, via personalisation, a profile — while the cleanest way to escape both is to start paying money. That’s a coherent business, but it’s also the quiet conversion of a free product into a funnel, and it belongs in the same story as every other way an AI plan is being re-priced after you committed to it.

What to actually do

If you use ChatGPT in one of the affected markets, the practical steps are undramatic:

• Know which tier you’re on. Ads reach Free and Go users only; Plus, Pro and Enterprise stay ad-free. If you already pay, nothing changes.
• Turn off personalisation if you value privacy over relevance. It won’t remove the ads, but it limits the profiling behind them — a real, if partial, win.
• Treat the paid tier as the ad-removal product it now is. If ads bother you on principle, that’s what the subscription buys; decide whether it’s worth it rather than expecting a free switch.
• Be sceptical of product recommendations inside answers. For now ads are labelled and separate. Keep noticing where the label is, and treat any purchase suggestion woven into an answer with the same caution you’d give a sponsored search result.

OpenAI has done this about as decently as an ad launch can be done: no selling your data, no handing advertisers your chats, ads that are labelled and kept apart from the answers, and paid tiers left clean. That is worth acknowledging, and it sets a bar the rest of the industry should be held to. The fixable fault is the smaller one — an “opt-out” that opts you out of relevance rather than ads, and a free tier that now quietly runs on your attention. The bargain isn’t outrageous. It’s just no longer the one on the box, and Europe, of all places, is where that difference will be tested.