OpenAI announced on 18 August that ChatGPT Ads is expanding into 31 European countries. Ads appear on the Free and Go plans. Plus, Pro and Enterprise stay ad-free.
If your team does research on the free tier, and in most organisations that have not bought licences they do, then the answers coming back are now on a surface that somebody can pay to influence. The paid tiers are the ones that are not.
That is the whole story, and it is a procurement question rather than an ethical one, which is what makes it worth acting on.
What was announced
This is the largest expansion of the product so far. It follows six months of a United States pilot plus eight further markets.
The countries named in the announcement include Germany, France, Spain, Italy, Sweden, Norway, Denmark, the Netherlands and Austria. Those are examples rather than the full list of 31, so if you need to know whether your own market is included, check the announcement rather than assuming from the names given. I would particularly avoid assuming about the United Kingdom either way.
Access initially runs through OpenAI’s own ads team, agency partners and technology partners, with a self-service Ads Manager coming later. Tens of thousands of marketers are already advertising on ChatGPT.
One thing the announcement does not describe is what an ad actually looks like inside an answer, or how it is labelled and separated from the response. That detail is not published, so anybody describing the format confidently right now is guessing.
Why this is not an outrage piece
I want to be clear that OpenAI is not doing anything underhand here. The tier split is stated openly. Free plans carry ads, paid plans do not, and that is a normal and long-established way to fund a service that costs a great deal to run.
The interesting part is not the ethics. It is that a decision most organisations have been deferring just acquired a concrete consequence.
The decision people have been deferring
Here is the shape of it in most places I talk to.
The organisation has not bought AI licences, or has bought a handful for a pilot. Meanwhile a good proportion of staff have signed up for a free account themselves, and they use it for real work. They summarise documents with it. They ask it to explain an unfamiliar regulation. They use it to research suppliers, competitors and markets before a meeting.
None of this appears in any procurement record. It is not shadow IT in the old sense of somebody installing software they should not have. It is people using a free tool because it helps and nobody told them not to.
Until now, the argument for buying licences has been about capability and data handling: better models, longer context, and a contractual position on what happens to what you type. Those are good arguments and they have been comfortably ignorable, because the free version mostly worked.
Advertising changes the character of the argument. The question is no longer only what the tool can do or where your text goes. It is whether the information your team is working from has a commercial interest attached to it.
What is actually at risk, stated carefully
I do not want to overstate this. An ad appearing alongside an answer is not the same as an answer being corrupted, and there is no evidence published that suggests the responses themselves are for sale.
But consider where this lands in practice. Somebody on your team asks for a comparison of suppliers in a category. Somebody else asks which tool is best for a job. Somebody asks for the leading options in a market you are entering.
Those are commercially valuable questions, asked at exactly the point where a decision is forming. That is what advertising has always paid for, and it is why search advertising became one of the largest businesses in the world. The mechanism is not sinister. It is simply worth knowing that it is now present in one tier of a tool your team may be using for exactly those questions, and absent in another.
The version of this that is easy to act on
The useful thing about a procurement framing is that it does not require anyone to agree about AI in general.
You do not have to win an argument about whether these tools are good, or whether they should be regulated, or whether the free version is acceptable in principle. You have to establish one fact and make one decision.
The fact: which tier is your team actually using. My strong suspicion is that most managers do not know, and that the honest answer in a lot of organisations is a mixture, with the split following whoever happened to expense something.
The decision: whether the work being done on the free tier is work you are comfortable having done there, now that the tier has a commercial layer the paid ones do not.
For some work the answer will be yes without much thought. Drafting an internal email does not require a clean information supply. For anything that feeds a purchasing decision, a supplier comparison or a market view, the answer deserves more attention than it has been getting.
One question this week
Find out which tier your team is actually using. Most people do not know.
That is the whole task, and it takes one conversation. Everything else follows from the answer, and you will probably learn something about how much AI is already inside your organisation while you are asking.
Source: ChatGPT Ads expands across Europe, OpenAI, 18 August 2026. The announcement names example countries rather than the full list of 31, and does not describe ad format or labelling.