OpenAI emailed users in Europe on 15 August 2026 to confirm that advertising will start appearing inside ChatGPT for people in the European Economic Area and Switzerland later this month. The message, timestamped 08:29 that morning, said the Free and Go plans will carry ads, while Plus, Pro, Enterprise, Business and Education plans stay ad free. PPC Land reported the email the same day.
This is the user side of ChatGPT advertising, and it is a different question from where advertisers are allowed to buy. Relevant Audience has already covered the advertiser-facing product changes: oCPC bidding, advanced matching, budget caps and the widening of health and finance advertiser eligibility. This announcement is about where the ads now show up, on which plans, and under what rules.
What OpenAI told European users on 15 August
The email OpenAI sent at 08:29 on 15 August 2026 confirmed that ads will begin appearing in ChatGPT for users in the European Economic Area and Switzerland later this month. It did not give an exact date. "Later this month" is the entire timing detail in the announcement, so anyone planning around it is planning around a window of roughly two weeks rather than a launch day. PPC Land, which reported the message on 15 August, carried the same wording.
The scope statement is worth reading literally. OpenAI named the European Economic Area and Switzerland. The source does not list the individual countries inside that grouping, and it does not say whether the switch flips for every market on the same day or in stages. Treat the geography as "EEA plus Switzerland" and nothing more precise than that until OpenAI publishes a date.
Free and Go carry the ads, the paid tiers do not
OpenAI said in the 15 August email that ads will appear on the Free and Go plans. Plus, Pro, Enterprise, Business and Education plans remain ad free. That is a clean split between the consumer entry tiers and everything a company is likely to buy for its staff.
The practical consequence for anyone trying to picture the audience is that the people who will see sponsored slots in ChatGPT are, by definition, the people who have not paid to remove them. Enterprise and Business seats are excluded, so the corporate desktop is not part of this surface. If your buyer is a procurement lead using a company-issued Enterprise account, this rollout does not touch them at all. If your buyer is a consumer or a small business owner on the Free or Go plan, it does.
Where the ad sits and how it is labelled
According to the same OpenAI message, ads render below the response text, are labelled "sponsored" and are visually separated from the answer. Three separate design decisions are packed into that sentence, and each one has a consequence.
- Placement below the response means the answer is generated first and the ad is appended after it. The ad is not woven into the sentence the model produces.
- The "sponsored" label is explicit, which is what regulators generally expect of paid placement inside an editorial or informational surface.
- Visual separation is stated as a design property, not merely as a label. OpenAI is drawing a boundary between the generated answer and the paid slot underneath it.
What the source does not describe is the creative format itself. There is no detail in the announcement about how many ads can appear under one response, what an ad unit contains, or whether it can carry an image. Those are open questions.
The launch model is contextual, not personalised
OpenAI stated that "to begin, ads will not be personalised". Ad selection uses only the topic of the current conversation, general location and device type. Personalised advertising requires a separate opt-in before it can be activated.
Read that carefully, because it describes an older kind of ad buying than most performance marketers work with day to day. There is no audience list, no behavioural profile, no cross-session history feeding the selection at launch. The strongest signal available is the subject of the message the user just sent. That makes ChatGPT ads, at launch, closer to classic contextual placement than to the audience-based buying that has dominated the last decade.
The phrase "to begin" also does real work. It signals that the non-personalised model is a starting position rather than a permanent design, and OpenAI has already described the mechanism by which it changes: a separate opt-in from the user. The source does not say when that opt-in will be offered, how it will be presented, or what share of users OpenAI expects to accept it.
What advertisers get back, and what they do not
OpenAI said advertisers "only receive overall information about how their ads perform, such as total views or clicks". That is the measurement contract at launch, and it is a narrow one.
Total views and total clicks are aggregate counts. Nothing in the announcement describes user-level reporting, audience breakdowns, conversation-topic breakdowns, or a path from an ad impression back to an identified person. For a marketer used to slicing performance by audience segment, this is a reporting surface closer to a print or out-of-home buy than to a self-serve social platform. You will know that the ads ran and roughly how they performed in aggregate, and the announcement gives no basis for expecting more than that at the start.
That constraint follows logically from the non-personalised model. If selection does not use a profile, there is no profile to report against. It is worth planning measurement on the assumption that attribution for this channel will have to come from your own analytics rather than from the platform, at least while the contextual model holds.
Who is excluded from seeing ads
OpenAI's email set out three exclusions. Users under 18 are excluded, and OpenAI said they are identified both through account age data and through behavioural prediction. Temporary Chat sessions are excluded. The Atlas browser is excluded.
The under-18 method is the most interesting of the three, because it is not purely a declared-age check. Behavioural prediction means OpenAI is inferring likely minors from how an account is used, not only from what the account says its birth date is. The source does not describe the signals behind that prediction or its accuracy.
The Temporary Chat exclusion aligns the ad system with the privacy promise of that mode: a session that is not retained also does not get monetised. The Atlas exclusion means the browser product stays clean of this ad surface for now.
Content the ads are kept away from
Health advertising, mental health advertising and political advertising are prohibited near sensitive content, according to the OpenAI message. That is a category-level restriction sitting on top of the placement rules.
The source does not define what counts as sensitive content, and it does not say whether the restriction is enforced by classifying the conversation, the advertiser, or both. For advertisers in regulated categories, the honest reading is that ChatGPT is treating those verticals as restricted next to sensitive conversations, and the detail of how that is judged is not public in this announcement.
The rollout order so far
The European launch is not the first. The dates below are the ones stated in the source, and they show a product that has been shipping market by market since June 2026.
| Market | Status per the source |
|---|---|
| United Kingdom | Live from 6 June 2026 |
| Japan and South Korea | Live from July 2026 |
| European Economic Area and Switzerland | Announced 15 August 2026, starting later this month, no exact date given |
Two things follow from that sequence. First, the pattern is a staged geographic rollout rather than a single global switch. Second, Japan and South Korea going live in July 2026 means the product has already reached Asia Pacific. The source says nothing about Thailand, so nothing here should be read as a Thai launch date, but the APAC precedent is a fact rather than a guess.
Who regulates this in Europe
The Irish Data Protection Commission is the lead supervisory authority for OpenAI Ireland Limited. That is the single most useful line in the announcement for anyone trying to work out where scrutiny of this rollout will come from.
A lead supervisory authority is the regulator that takes primary responsibility for a company's cross-border data processing in the European system. Naming the Irish DPC tells you which door questions about the ad system's data handling will be knocking on. The source does not report any regulatory decision, investigation or approval attached to this launch. It reports who the lead authority is.
What the announcement did not say
Being precise about the gaps is as useful as repeating the facts.
- No exact launch date. Only "later this month".
- Nothing about Thailand, or about any market outside the ones named.
- No pricing, no ad formats, no description of how the inventory is bought or which advertisers are eligible in the EEA.
- No statement about how many ads can appear per conversation or per session.
- No detail on whether the presence of an ad changes the generated answer above it, beyond the statement that ads sit below the response and are visually separated.
- No timing or mechanism detail for the personalised-ads opt-in beyond the fact that it is separate and required.
What this means for Thai marketers
Nothing in this announcement changes anything inside a Thai account today, and the source does not mention Thailand. What it changes is the planning assumption. A staged rollout that has already covered the United Kingdom, Japan and South Korea before reaching Europe is a product being pushed out market by market, and Japan and South Korea show it is not a Western-markets-only launch.
Two pieces of preparation are worth doing while the surface is still contextual. The first is topical: if selection at launch runs on the topic of the conversation, then being the brand associated with a topic matters more than owning an audience list. That is the same discipline as being present in AI answers in the first place, which is why ChatGPT SEO and the broader work of generative engine optimisation sit next to this rather than being replaced by it. Paid placement under an answer and organic presence inside the answer are different slots, and both are decided by topic.
The second is measurement. If the reporting an advertiser receives is limited to totals such as views or clicks, then any serious read on this channel will have to be built from your own side: clean landing pages, consistent tagging, and analytics you trust. That work is not new, but a channel with thin platform-side reporting raises the cost of not having it.
Questions people are asking
Is this live in Thailand?
The source does not say anything about Thailand. It reports the ad rollout for the European Economic Area and Switzerland, and lists the United Kingdom from 6 June 2026 and Japan and South Korea from July 2026 as the markets that came before. There is no Thai launch date in this announcement.
Do I have to do anything right now?
No. Nothing in this announcement requires an action from an advertiser outside the affected markets. It changes where ChatGPT ads are shown to users, not how any existing account is configured.
Will my ChatGPT Plus or Enterprise account show ads?
No, according to OpenAI's email. Plus, Pro, Enterprise, Business and Education plans stay ad free, and ads run on the Free and Go plans.
Are these ads targeted at me personally?
OpenAI stated that "to begin, ads will not be personalised". Selection uses the topic of the current conversation, general location and device type, and personalised advertising requires a separate opt-in before it is activated. The wording "to begin" indicates this is the starting configuration rather than a permanent one.
When exactly do the ads start in Europe?
The source does not give an exact date. OpenAI's email said later this month, meaning later in August 2026, and no more precise timing was published in the announcement.
Where to go from here
The detail in this piece comes from the OpenAI email reported by PPC Land on 15 August 2026. If you want a view on how your brand shows up inside AI answers, and how a contextual ad surface underneath those answers might sit alongside that, Relevant Audience is happy to walk through it with you.
