OpenAI has been doing ads in ChatGPT in the US since February and just launched in Australia, New Zealand, and Canada.
This surprised people who’d never encountered Sam Altman. Here’s Altman at Harvard Business School in May 2024, answering an audience question about whether he’d put ads into ChatGPT: [YouTube]
You’ll see us do a lot more to make the free tier much better over time. And I’m interested in figuring out how we bring the equivalent concept to the API. But I kind of think of ads as like a last resort for us for a business model.
It’s 2026, and OpenAI’s getting a bit last-resorty.
OpenAI’s business numbers for ChatGPT have never worked, ever. We’ve been yelling about this since 2024, when Altman was saying “no ads.”
OpenAI was spending $2.35 for each dollar of revenue then. It’s only gotten more expensive since. And OpenAI is starting to run out of other people’s money.
Near the end of 2025, OpenAI hired on a whole pile of people who’d most recently worked at Meta. People who were quite well acquainted with Meta’s stupendous digital advertising engine: [Information, archive]
staff have angsted over the prevalence of people at OpenAI who previously worked for Meta, and over whether OpenAI will become more like the social media and digital advertising giant.
Sam wishes OpenAI would become more like Meta. In November 2025, developer Tibor Blaho found a pile of ad code inside the ChatGPT Android app: [Twitter, archive]
ChatGPT Android app 1.2025.329 beta includes new references to an “ads feature” with “bazaar content”, “search ad” and “search ads carousel”.
In December, OpenAI floated the idea of ads in the chatbot output a bit more aggressively: [Information, archive]
Employees have discussed ways to tweak AI models to prioritize sponsored information in ChatGPT’s responses when users ask relevant queries, a person familiar with the discussions said. For instance, a Sephora-sponsored beauty product ad could appear when a user is searching for mascara recommendations.
… some employees feel an ads push would be counter to the company’s loftier goals of achieving artificial general intelligence, or when AI can surpass human performance in a variety of tasks.
Ads? Or Roko’s basilisk? Tricky choice. OpenAI went for the ads. The company announced in January it was going to start ad testing. On 9 February, the ads went live. [OpenAI; OpenAI]
ChatGPT shows the ads in a box below the chat. The ads only show up on the free and “Go” levels. The $20-a-month subscribers won’t see the ads. Yet.
OpenAI thinks the ad programme will be the biggest thing ever. They claim their ads will pull in $102 billion by 2030: [Information, archive]
OpenAI expects advertising to generate about $2.4 billion in revenue this year and to quadruple next year, to nearly $11 billion, according to financial forecasts from the first quarter, which haven’t previously been reported.
… In 2030, OpenAI expects ads to generate about $102 billion, or 36% of its total revenue for that year.
OpenAI is just making up imaginary future numbers here. For comparison, Meta made $196 billion in ad revenue in 2025. That’s across all of Facebook, Instagram, Threads and WhatsApp. Netflix took in $1.5 billion in 2025 from TV ads. [Meta; Netflix, PDF]
OpenAI’s ad prices start at $60 per 1,000 impressions. That’s a premium rate. It’s the sort of price you pay for ads on a live NFL football game. I do not believe free OpenAI users are as premium as that. [FT, archive]
The ads are not so attractive to ad buyers because OpenAI’s advertisers can’t show their clients if OpenAI ads even worked. Also, the ads are weirdly hard to buy: [Information, archive]
OpenAI hasn’t yet offered marketers any automated way to buy ad space. Buyers have had to rely on making phone calls and sending spreadsheets and emails to OpenAI representatives, one ad executive said. More importantly, advertisers found it hard to tell whether the ads were paying off.
That’s partly because OpenAI hasn’t provided advertisers with much information beyond how often people have viewed and clicked on the ads. That’s a contrast to established digital media firms and ad technology companies, which offer advertisers a clearer picture of the type of audience seeing an ad and whether it drove business.
That is, the advertisers want the sort of intrusive personal data they expect from Facebook and Google and our 900 trusted partners.
OpenAI understands that sort of personal data leakage is a formula for yet more bad press. But also, they really need the money. OpenAI is running out of ready cash. OpenAI is flailing about looking for money anywhere.
So OpenAI is, in due course, going to give your personal details to 900 trusted partners. And they will tweak the output of the chatbot to be a bit more advertiser-friendly. They can’t afford not to.
OpenAI cannot pump through ads nearly fast enough to make up the fantasy numbers they’re forecasting and beat their ghastly burn rate. This is not the one weird trick that just might work. The ads might make ChatGPT suck even more, though.
It’s pledge week at Pivot to AI! If Pivot brightens your day, please do put $5 into the Patreon. Tell your friends!
















