Having burned through about $8 billion of cash in 2025, OpenAI seems in desperate need of revenue. Now it’s jumping into what Chief Executive Sam Altman described in 2024 as the company’s last resort: showing ads in ChatGPT. Doing this so soon after that dismissal speaks to the financial pressures his company is under.1
Investors will cheer at the effort to beef up its top line and compete with ad king Google, which is preparing to run adverts in its chatbot too, and online advertising is one of the most successful businesses of all time. It has kept large swathes of the internet free and dynamic. But it has also been socially damaging, incentivizing rage bait, misinformation and other toxic byproducts of our smartphones. OpenAI says it won’t go the same way. I have doubts about a company that has taken a few U-turns in its lifetime.
OpenAI is ruling out the more obvious downsides for users of its free and $8-a-month ChatGPT Go tiers who will start seeing ads. Advertisers won’t influence the chatbot’s answers but rather post banners with images at the bottom of the screen. Ask a question about cooking a Mexican meal, for instance, and it’ll display an ad for hot sauce. And what of privacy? The company says it won’t sell users’ personal data and will keep ChatGPT conversations private.
There’s a decent likelihood of OpenAI keeping both those promises in the short term. Banner ads are as old-fashioned as they are irritating, but they still exist on apps like Meta Platforms Inc.’s Facebook and LinkedIn because they work.
As for scooping up personal information, Meta showed it could thrive with less data thanks to the workarounds it figured out after Apple effectively killed the ad cookie in 2021, allowing millions of people to block apps from tracking their data. Conventional wisdom at the time was that Mark Zuckerberg’s business — which derives about 98% of its revenue from advertising — would be dealt a terrible blow. Instead, Meta successfully adapted by using AI-driven modelling to help it place ads and not just personal data.
It is OpenAI’s final pledge that looks harder to maintain. The company promises it won’t “optimize for time spent in ChatGPT,” and by this it means it won’t prompt its model to keep users engaged for as long as possible, and it won’t try to maximize the time that people’s eyeballs are spent on ChatGPT looking at ads. This pledge won’t only be hard to stick by, it’s also difficult to measure.
Social media’s ability to make money from ads is so powerful that in some cases even a small uptick in “engagement” can lead to outsized financial benefits. When Meta reported an 8% rise in user engagement in the second quarter of 2025, for instance, that came with a 22% jump in ad revenue. So any minor tweaks that OpenAI makes to ChatGPT to juice engagement might not be very noticeable, but they could still boost revenue. Given that the company is entirely opaque about the mechanics of its AI models (despite its name), independent researchers would struggle to detect such changes, too.
OpenAI also doesn’t have to do much to make ChatGPT more alluring, just turn up the charm a little. Large language models are already known for their tendency to flatter users, praising their prompts as “fascinating” or “thoughtful” for example, and for using anthropomorphic language to make themselves sound more human, with phrases like “I’m listening” or “I’m concerned.”
Beefing up those features even slightly would be difficult to detect from the outside, and easily justifiable within OpenAI’s product-development meetings as part of an effort to make ChatGPT more user friendly. But they could increase time spent on the app, bringing with it many of the societal and health problems of keeping people glued to their phones, issues that have been underscored by several recent lawsuits against Altman’s firm.
OpenAI’s recent recruitment may be instructive. It spent much of 2025 hiring staff from Meta, grabbing expertise from a company that wrote the book on optimizing algorithms for engagement. As of October, about 630 former Meta employees worked at OpenAI, making up about 20% of its roughly 3,000-strong workforce, according to tech-news site The Information.
And they do have a lot of eyeballs to go for. ChatGPT is being used at massive scale, with 900 million people chatting to the bot every week, according to OpenAI’s latest figures. Data from web analytics firms suggest the average session runs to around 15 or 20 minutes, with many users returning several times per day. That’s a tantalizing metric for advertisers.
With OpenAI’s finances stretched, it’s hard to see Altman resisting the gravitational pull of the ad model and avoiding any of the compulsive mechanics that drive today’s mobile games, social media and websites. After all, he first founded his company as a non-profit to benefit humanity and then turned it into a commercial enterprise, was ousted by his board for being flexible with the truth and has now pivoted to ads a year after dismissing them. His plans tend to shift under pressure.
Instead of purely designing ChatGPT to be helpful, he’s on track to optimize it to be habit forming, all while inadvertently targeting the most vulnerable: those who can’t afford a $20 a month subscription, including many young people. In an ideal world Altman will resist the warped incentives of online advertising. But history shows that ads reshape internet platforms. There’s little reason to think ChatGPT will prove the exception.
1. Financial author Sebastian Mallaby recently predicted the company would run out of money in the next 18 months and be absorbed into Microsoft Corp.
A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Parmy Olson
