ChatGPT already has ads: will users accept to pay with their attention?
The decision that no one expected (or that we all saw coming)
OpenAI has started showing ads to users of its free and Go plans, which is the $8/month plan it launched in January. If you pay for Plus, Pro, Business, or Enterprise, for now, you don’t get it. I say for the moment because we know how these things usually end up.
Sound familiar? It’s the same thing Google said when it put ads in its search engine over 20 years ago. And look where we are now.
The decision that no one expected (or that we all saw coming)
OpenAI has started showing ads to users of its free and Go plans, which is the $8/month plan it launched in January. If you pay for Plus, Pro, Business, or Enterprise, for now, you don’t get it. I say for the moment because we know how these things usually end up.
The ads are labeled as “sponsored” below the responses. They are separate from the generated content, and OpenAI swears that advertising has nothing to do with what ChatGPT answers you. That the advertising and text generation systems go their separate ways.
Sound familiar? It’s the same thing Google said when it put ads in its search engine over 20 years ago. And look where we are now.
The numbers behind this decision
Okay, before we judge, let’s look at the numbers. Because the underlying question is not whether the ads are annoying (which they obviously are), but whether OpenAI can afford not to run them
According to the Financial Times, only 5% of the 800 million users pay for ChatGPT. To give you an idea: that’s 760 million people using state-of-the-art AI infrastructure without dropping a euro. Totally free. And that infrastructure costs an awful lot to maintain. OpenAI has not yet made a profit, and its investors have billions in it.
Their internal documents project that monetizing free users will generate them about $1 billion this year, escalating to nearly $25 billion by 2029. With an IPO planned for the end of 2026, they need to prove that the racket is sustainable.
And how much do you charge? The CPM (cost per thousand impressions) is $60. To put it in perspective, Meta charges three times less. In addition, the barrier to entry for advertisers is high: a minimum investment of $200,000. The first to sign up have been WPP, Omnicom, and Dentsu with retail, software, and travel brands.
Sam Altman: from hating ads to embracing them
Here’s what struck me most about the whole thing. During the early years of ChatGPT, Sam Altman was very clear on the subject: he said he found the combination of advertising and artificial intelligence “extremely disruptive.” Just like that. In October 2024, he was still calling ads a “last resort” for his business model.
Eighteen months later, the last resort arrived. What changed? Well, the usual: OpenAI’s transformation from a nonprofit to a for-profit company (completed in October 2025) and the pressure of the IPO. Financial reality has a habit of rewriting principles.
The resignation that says more than a thousand communiqués
On the same day that the announcements went live, OpenAI researcher Zoë Hitzig resigned. She was quite blunt: she said she didn’t trust OpenAI to protect what she described as “the most detailed record of private human thought ever assembled.”
And let’s see, think about it for a moment. ChatGPT knows what you’re thinking about, what you’re worried about, what you’re looking for at three in the morning when you can’t sleep. That level of user intimacy has not been available to Facebook or Google at their best. Hitzig warned that OpenAI is riding an economic engine with“strong incentives to bend its own rules” as the ad business grows. Coming from someone who used to work on the inside, that’s no small thing.
Anthropic laughs at the Super Bowl
As if OpenAI’s controversy wasn’t enough, Anthropic took advantage of the Super Bowl to release a series of spots mocking the idea of putting ads into an AI. They showed absurd situations of ads interrupting conversations, with the tagline, “Ads are coming to AI. But not to Claude.”
Altman’s reaction was quite revealing. He called Anthropic’s ads “dishonest” and the company “overbearing.” When competitors strike a nerve like that, it’s usually because they’ve hit where it hurts.
Can ads skew responses? The debate is open
Here’s the crux of the matter. OpenAI insists that advertising and responses are handled by separate systems. But experts are not so sure.
Miranda Bogen, who heads the AI Governance Lab at the Center for Democracy & Technology, explains it well: even if platforms don’t share your data directly with advertisers, business models based on targeted advertising create dangerous incentives. It’s the same old dynamic, but applied to something much more intimate than a social media feed.
And then there’s something that I find even more worrying: the concept of GEO, “Generative Engine Optimization,” which we’ve discussed in other articles. It works like SEO, but for AI. Instead of optimizing content to appear first on Google, the aim is to have certain products mentioned favorably in the model’s responses. Not as a visible advertisement, but embedded within the response, disguised as an expert recommendation. If this scales (and I have little doubt that it will), we are going to have a big problem distinguishing between real advice and camouflaged advertising.
User reaction: what was to be expected
On Reddit, 68% of comments were negative. On X, discussions exceeded 10 million views and the tone was mostly skeptical.
But there is one thing that I find even more significant. Weeks before the official launch, users had already caught promotional messages sneaking into unrelated conversations. People talking about programming suddenly saw specific suggestions. Even paying subscribers reported it. OpenAI had to quickly disable them, and Mark Chen, its director of research, admitted on X that “anything that looks like an ad needs to be handled with care, and we fell short.”
When your own team publicly acknowledges that they “fell short”… that’s bad news.
What if I say I don't want ads?
OpenAI gives you options, yes. You can disable personalized ads. But beware of the fine print: if you disable them, you will still see ads based on the context of your current conversation. What disappears is the personalization based on your previous chat history. To remove ads altogether, upgrade to a higher-tier plan.
In other words: either you pay with your wallet, or you pay with your attention and your data. The choice is yours.
This isn't the end of it
Google has already confirmed that it plans to introduce advertising into Gemini in 2026. If the two giants of conversational AI adopt this model, it will become the standard. And that will profoundly change the relationship we have with these tools.
The value of ChatGPT has always been that feeling that it gives you objective answers. No agenda. You ask questions and it answers to the best of its ability. The day users start thinking, “Is it recommending this to me because it’s the best or because
My thoughts
I think the question we should be asking is not whether the ads are good or bad. It is another: will users accept them?
OpenAI has built something that 800 million people use. That costs a fortune to maintain, and they need to monetize, I totally understand that. But the execution is leaving something to be desired. The pre-launch screw-ups, Hitzig’s resignation, and especially the speed with which Altman went from “I find the ads disturbing” to introducing advertising partners to society… all of that creates a sense that priorities within the company have changed in a major way.
And when the priorities of the company managing your personal AI assistant change, perhaps yours should change too.
What do you think, do you think it’s reasonable for ChatGPT to show ads in exchange for being free, would you rather pay to not be bothered, do you trust that ads will never influence what you get back?
Leave me your comments, I’d love to read them. This debate has only just begun.
Have a good week!
Article sources: OpenAI (official blog), TechCrunch, Financial Times, Euronews, WinBuzzer, Marketing AI Institute, Center for Democracy & Technology, Proton Blog, among others.
