“We’ve seen a lot of stuff around ads from some of the other players. Anthropic is not a player that works like, or needs to work like, that. We just sell things to businesses, and those things directly have value. We don’t need to monetize a billion free users. We don’t need to maximize engagement for a billion free users because we’re in some death race with some other large player.”
—Dario Amodei, the C.E.O. of Anthropic, taking a shot at OpenAI during an interview in Davos earlier this week.

Runner-up: “We’re not going to kill all our employees tomorrow because of A.I.” —Jamie Dimon, the C.E.O. of JPMorgan Chase, also at Davos.

And now for the main event…

Two years ago, while speaking at Harvard, OpenAI C.E.O. Sam Altman confidently declared that advertising was “a last resort for us as a business model.” At the time, OpenAI was on track to lose something like $5 billion—and the company’s annual cash burn has only accelerated in the intervening years. By the end of 2025, analysts projected that it could lose double that amount, with total capex forecasted to rise above $100 billion per year through the end of the decade. Other analysts have been even more bearish: HSBC expects OpenAI to face a $207 billion funding gap by 2030; in December, an analyst at Deutsche Bank said that “no start-up in history has operated with expected losses on anything approaching this scale.”

In retrospect, it was probably inevitable that OpenAI, which claims to have roughly 800 million weekly users, would eventually start selling ads. Google and Meta generate hundreds of billions of dollars every year from their advertising businesses. So it was hardly shocking when the company announced, last week, that advertising was officially coming to ChatGPT. In my conversations with industry sources, most began with some variation of, “I’m not surprised at all.”

Here’s what we know about the ad rollout so far: Beginning in February, OpenAI will start testing ads for adults using the free product, as well as the Go tier—ChatGPT’s lowest-cost subscription. When the ads start appearing, OpenAI says they’ll look like a pop-up at the bottom of the screen—clearly marked as an ad and separated from the A.I.-generated output above it. In its announcement, OpenAI also assured users that ads would not influence output; that the company would “always offer a way to not see ads in ChatGPT” (presumably by paying for a more expensive tier); and that it would not sell user data to advertisers. We’ll see how long that lasts.

A sample of what this might look like. Courtesy of OpenAI

An OpenAI spokesperson told me that this promise covers all types of user data, though this person also said that advertisers will have access to analytics that show aggregated performance. But it’s hard to imagine that the company, having cracked open this particular Pandora’s box, won’t eventually pull most available levers. OpenAI knows a lot about its users, which it can use to optimize ad targeting. And Altman has burned through a lot of capital from investors who will eventually require a return. “Beyond charging for subscriptions, the only other way to monetize ChatGPT is through advertising,” D.A. Davidson analyst Gil Luria told me.

Obviously, launching an advertising business doesn’t mean that OpenAI is going to abandon its pursuit of subscription revenue—even though, as of July, only about 5 percent of its users were on a paid tier. The Information recently reported that OpenAI plans to increase its average revenue per user from $2 this year to $15 in 2030. But observers like Pedr Howard, an executive V.P. at Ipsos—which works with OpenAI on brand strategy—are skeptical that individual (as opposed to enterprise) subscriptions will ever become a meaningful revenue driver for the consumer-facing company. “We’ve tested a lot of ads for L.L.M.s over the years—there’s going to be plenty of A.I. ads in the Super Bowl—and haven’t seen one yet where people have really been convinced by the proposition,” he said.

A digital platform serving contextually relevant ads based on data analytics is nothing new. But the OpenAI pivot feels different, given the particular dynamics of ChatGPT. Unlike the passive scrolling experience on Instagram, Google, or YouTube, chatbots offer a much more active, anthropomorphic experience—engendering a sense of trust that is easy to betray. “What’s happening here isn’t just advertising inside a product, but advertising embedded inside interaction, advice, and decision support,” advertising scholar Dr. Jihoon Kim told me. “In search or social media, ads respond to inferred interests. In ChatGPT, ads could respond to expressed intent, people explaining problems, asking for guidance, and thinking out loud in natural language. That shift toward contextual intimacy is where both the opportunity and the risk come from.

The opportunity, according to Chris Shilney, a vice president at marketing firm Walker Sands, is “highly contextual, intent-rich” advertising, the impact of which depends on maintaining user trust (a fine line). The risk is that the rollout will “erode trust in a system users increasingly treat as a cognitive partner rather than a media platform,” said Kim. “It is this psychological closeness that makes the shift both powerful and delicate from a design and trust perspective.”

Decades of observing consumer behavior on the internet give us a decent idea of how users might respond. Paddy Harrington, a security and risk analyst at Forrester, told me that roughly a third of online users seem to accept advertising as a cost of admission to a service; another third hate advertisements but decide to use services with them anyway; and another (growing) third tend to abandon these platforms for ad-free alternatives. How this shakes out may come down partly to design. “If they put it outside of the chat flow, it’s annoying, but it’s tolerable,” Harrington said. “But at the end of a chat line, especially for people who use it as that emotional connection at some level, it feels a little slimy.”

Granted, OpenAI said that ads won’t initially appear next to “sensitive or regulated topics like health, mental health, or politics.” But a quick scan through OpenAI’s subreddit indicates that some users, at least, are already thinking about switching to alternatives like Gemini and Claude. To that end, Google DeepMind C.E.O. Demis Hassabis said this week that Google doesn’t plan on bringing ads to Gemini. “It’s interesting they’ve gone for that so early,” Hassabis told journalist Alex Heath. “Maybe they feel they need to make more revenue.” (It’s worth pointing out that while Gemini doesn’t serve ads inside Gemini, it does collect user data to serve ads elsewhere.)

No matter how this goes for OpenAI, it seems likely to open the A.I. advertising market across the board. Nic Baird, a co-founder of Koah, a startup focused on bringing advertising to A.I. applications, told me that the biggest objection he gets from publishers is that they don’t want to be first. But now that OpenAI is starting to embrace ads, the game has changed. Computational linguist Maria Sukhareva, for one, expects many A.I. developers will eventually resort to ads, whether they’re hyperscalers or not. “Google has already been considering for a while to introduce ads to the bot,” she told me. “The difference is that for OpenAI, it might eventually be the question of surviving longer; for Google, it will just be another source of income.”

That’s all for today. I’ll see you next week.

Ian

Need help? Review our FAQ page or contact us for assistance. For brand partnerships, email ads@puck.news.

You received this email because you signed up to receive emails from Puck, or as part of your Puck account associated with {{customer.email}}. To stop receiving this newsletter and/or manage all your email preferences, click here.

Puck is published by Heat Media LLC. 107 Greenwich St., New York, NY 10006