OpenAI is testing advertising in ChatGPT for the first time in the coming weeks – only in the US, only for free and Go users, only at the end of responses. Plus, Pro and Enterprise remain ad-free. That sounds harmless, but it’s the beginning of a new era: AI assistants are becoming advertising platforms. The question isn’t whether, but how invasive.
It was only a matter of time. ChatGPT costs OpenAI billions – for computing power, infrastructure, personnel. Revenue from Plus subscriptions (20 dollars), Pro (200 dollars) and Enterprise contracts doesn’t cover it. So now comes what eventually comes to every successful tech product: advertising. OpenAI will test adverts in ChatGPT for the first time in the coming weeks, initially only in the US and only for adult users of the free version and the new “Go” tier (8 dollars per month). Plus, Pro, Business and Enterprise remain ad-free. The adverts appear as clearly marked, separate boxes at the end of a response and according to OpenAI shouldn’t influence the content of responses themselves. That’s the theory. Practice begins now.
Current Status: What We Know
As of January 2026, no advertising is live yet, but test launch “in the coming weeks” in the US has been announced. Affected are free accounts – logged-in, adult users in the US – and the new ChatGPT Go, a cheaper tier with memory and more images, but with advertising. There’s no advertising in Plus (20 dollars), Pro (200 dollars), Business and Enterprise. The aim: additional revenue to finance broader access – Free and Go – without raising prices of the more expensive plans. The logic is classic: those who pay get no advertising. Those who don’t pay become the product.
How the Advertising Is Built In
Placement in the interface is clearly defined: adverts are displayed at the bottom of the ChatGPT response in a separate block, separated from the actual model output. They’re marked as “Sponsored” or as advertising and visually set apart. That sounds harmless – no pop-ups, no interruptions, no adverts in the body text. Just a block at the end.
The trigger is contextual: content should match the current conversation – travel planning leads to hotel adverts, product questions to shopping links. During the test phase, no ads on sensitive topics: health, mental health, politics are off-limits. Interaction is optional: users can see why an advert was shown, click it away and explain why it’s disruptive or inappropriate. That’s user control – at least on paper.
Privacy, Targeting and the Question of Independence
OpenAI emphasises “Answer Independence”: advertising doesn’t control responses. Responses are optimised for usefulness independently of adverts. That’s important, because if responses are influenced by advertisers, ChatGPT is no longer an assistant but a sales channel. OpenAI promises: “We don’t sell your data to advertisers.” Conversations shouldn’t be sold to advertisers. Targeting is based primarily on conversation context – intent – and possibly limited personalisation. Aggregate statistics – impressions, clicks – are given to advertisers, but not individual conversation content.
User control is planned: settings to deactivate personalisation for advertising, and the ability to delete data used for this purpose. And there should always be a paid, ad-free option – Plus, Pro, Business, Enterprise. The question is: how long does this boundary remain? How long does “Answer Independence” remain truly independent? And how transparent is what “limited personalisation” means?
Planned Formats and Expansion Stages
The first phase – test 2026 – is simple: static “Sponsored” blocks beneath the response, with link to product or service. No videos, no interactive elements, no pop-ups. Just text and link. That’s conservative, but it’s the beginning.
The prospect for later expansion stages isn’t yet official, but market observers expect conversational shopping ads: the user asks questions about a product directly in the conversation, compares options, prepares purchases – and the advert becomes part of the conversation, not just a block at the end. More strongly contextual, native formats instead of classic banners. That’s the logical next step, and it’s dangerous. Because when advertising becomes part of the conversation, the line between recommendation and advert blurs. And then “Answer Independence” becomes a promise that’s difficult to keep.
What Remains Open
There are no official details on pricing model – CPM, CPC, CPA, revenue share – only speculation about rather premium prices due to high intent signals. Anyone asking ChatGPT about a product has concrete interest. That’s more valuable than a banner on a website. No exact international rollout: timing for EU and Germany is unclear, only the hint that global expansion will follow US tests. No concrete self-service platform for advertisers or booking routes – whether direct, via agencies or partners like Reddit, remains open.
What This Means
OpenAI is testing advertising because it needs money. That’s legitimate. But it changes what ChatGPT is. ChatGPT was a tool – neutral, useful, on the user’s side. With advertising it becomes a platform – with interests, with partners, with economic incentives that don’t always align with users’ interests.
The question isn’t whether advertising is coming. It’s coming. The question is how invasive it becomes. Today it’s a block at the end of the response. Tomorrow it could be a recommendation within the response. The day after it could be a conversation sponsored by an advertiser without you noticing.
OpenAI promises transparency, control, independence. These are important promises. But they must be kept. Because when the assistant becomes an advertising medium, the question is no longer whether it helps. But: whom it helps.
When the assistant sells, it must be clear what it’s selling – and for whom.
