On 18 August 2026 OpenAI announced that ChatGPT Ads would expand to thirty-one European markets, and from the week of 24 August the ads started appearing in conversations across Italy, Germany, France, Spain, Austria, the Netherlands, Sweden, Norway and Denmark, among others. Most of the coverage framed it as a straightforward geographic rollout. It isn’t. What launched in Europe is a materially different product from the one American advertisers have been buying since May, and the differences sit exactly where budget decisions get made.
Two of them matter more than everything else. There is no self-serve buying in Europe yet, which means a European business cannot simply open an account and launch a campaign the way a business in Ohio can. And personalisation works differently across the European Economic Area, which means the signal set the auction draws on is thinner than the one behind the American benchmarks everyone is quoting.
If you are a US advertiser planning European expansion, or an agency briefing a client on where ChatGPT Ads fits in a multi-market plan, this is the part of the picture that the American guides cannot give you.
In brief
ChatGPT Ads launched across thirty-one European markets from the week of 24 August 2026, roughly six months after the US pilot began. Ads are shown only to users on the Free and Go plans; Plus, Pro and business tiers remain ad-free, as do accounts identified as belonging to under-18s. Self-serve buying through Ads Manager is not open in Europe: initial access runs through OpenAI’s Ads Solutions team, agency partners and technology partners, with self-serve indicated for later in the quarter. Personalisation based on conversation history requires explicit opt-in consent in the EEA, and users who decline still see ads, but contextual ones only. No published cost benchmark covers European campaigns.
How Europe got here
The timeline is worth having straight, because it explains why the European version looks the way it does. OpenAI signalled the arrival of advertising in January 2026 and launched the US test on 9 February for adult registered users on the Free and Go tiers. Canada, Australia and New Zealand followed in March. The beta Ads Manager arrived in spring, first with a lowered pilot minimum and cost-per-click bidding, then in May with self-serve access opened to US advertisers and the minimum spend requirement removed entirely. Over the summer the UK, Mexico, Brazil, Japan and South Korea came online. Europe arrived last, and arrived at scale.
By the time the European rollout happened, the platform itself had matured well beyond a test. Bidding, geographic targeting, custom audiences, conversion optimisation, the OpenAI Pixel, a Conversions API and third-party measurement integrations were all live features. The gap in Europe is not one of platform capability. It is a gap between what the platform can do and what a European advertiser can currently reach.
The access gap
In the United States, buying is direct. An advertiser registers, adds payment details, sets budgets, bids and pacing, uploads creative, submits for review and manages performance in the portal. That door is not open in Europe. Initial European access runs through OpenAI’s Ads Solutions team, through agency partners — the holding groups have been named in reporting on the launch — and through technology partners. OpenAI has indicated self-serve availability later in the quarter, without committing to a date.
The practical consequences are worth spelling out. Test sizes in Europe are, for now, effectively set by what an intermediary will take on rather than by the advertiser’s own appetite. The first wave of European performance data will come from larger, partner-managed accounts, which means it will skew toward categories and budget levels that are not representative of the market as a whole. And the small-and-mid-market experimentation that produced most of the interesting early US data simply has not started yet on this side of the Atlantic.
For a US brand planning European entry, that argues for sequencing rather than simultaneity. The American account can keep generating creative and context-hint learnings now; the European buy becomes a separate decision when self-serve opens, with the American learnings as a starting hypothesis rather than a template.
The personalisation gap
The second difference is more interesting strategically. In the EEA, personalisation drawing on conversation history and prior ad interactions requires explicit opt-in consent. Users who decline still see advertising, but it is contextual: driven by the conversation currently happening rather than by an accumulated profile.
That changes the composition of the signal behind ad selection. In the US, the system can lean on a layered set of inputs. In Europe, for a meaningful share of the audience, several of those layers are absent. The weight shifts onto relevance to the conversation in front of the model, which in turn shifts weight onto the two things the advertiser actually controls: how well the context hints describe the situations where the product genuinely helps, and how well the landing page matches the promise of the ad.
This is why the framing of ChatGPT Ads as “the keyword version of conversational advertising” is misleading everywhere, and particularly misleading in Europe. Context hints are not keywords, and treating them as a phrase list to be expanded is the fastest way to burn a European test budget. In a market where profile signal is thinner, describing the situation precisely is not an optimisation detail. It is the mechanism.
What the US benchmarks can and cannot tell you
There is now a reasonable amount of published US performance data, and it disagrees with itself. Recommended starting bids in the three-to-five dollar range for consumer categories sit alongside considerably higher figures for software and finance, while at least one agency has published its own account data showing a cost per click well under two dollars over a two-week window, with a conversion rate in the low single digits and a return on ad spend just under one and a half. Early pilot click-through rates around one per cent circulated widely and were then contradicted by later blended figures several times higher.
None of this is contradictory in a suspicious way. It is what a young auction looks like: pricing moves as advertiser demand fills available inventory, and the composition of who is bidding changes month to month. It does mean that quoting any single figure as “the ChatGPT Ads CPC” is indefensible, and quoting a US figure as a European planning input is worse, because it adds a currency, a market, a consent regime and an access model on top of the underlying volatility.
The usable approach is the one that works for any new channel. Start from unit economics rather than from a market average. Model a conservative case, a target case and a strong case, then replace each assumption with observed data as it arrives. Cost per click tells you what traffic costs; it does not tell you what a customer costs. If conversational context genuinely delivers users who have already articulated a problem, compared alternatives and set a budget, a more expensive click can still produce a cheaper acquisition. That claim is testable, and until someone tests it in a European market with European consent settings, it remains a hypothesis.
Who should be paying attention now
The categories most visible among early US adopters are the ones where users describe complex problems before deciding: B2B software, professional services, education, travel and financial services, and high-consideration ecommerce. Those are also, not coincidentally, among the most expensive categories on Google and Meta, which is what makes an emerging auction attractive to them in the first place.
For Europe specifically, the argument for engaging now is not that the channel is proven. It is that the preparation is transferable regardless of what the channel turns out to be worth. Getting conversion tracking into a state where a new source can actually be measured, mapping which conversations realistically precede a purchase in your category, and writing honest descriptions of where your product is genuinely the right answer — none of that is wasted work if ChatGPT Ads underperforms, and all of it is the difference between a structured first test and a repurposed Meta campaign when self-serve opens.
FAQ
Is ChatGPT Ads available in Europe? Yes, from the week of 24 August 2026, across thirty-one markets announced by OpenAI on 18 August.
Can a European business open a ChatGPT Ads account directly? Not yet. Access runs through OpenAI’s Ads Solutions team, agency partners and technology partners. Self-serve has been indicated for later in the quarter.
Who sees the ads? Users on the Free and Go plans. Plus, Pro and business tiers do not show ads, and neither do accounts identified as belonging to under-18s.
Does targeting work the same way as in the US? No. Personalisation based on conversation history requires explicit opt-in consent in the EEA; users who decline see contextual advertising only.
Do ads influence ChatGPT’s answers? OpenAI states that ads are kept separate from responses and that advertisers cannot alter or influence them.
Are there European cost benchmarks? Not yet. Published figures are in dollars and reflect US campaigns under a different access and consent model.
