OpenAI tests ChatGPT ads for US Free and Go users as financial pressure mounts
OpenAI is testing ads inside ChatGPT for logged-in Free and Go-tier users in the United States. Plus, Pro, Business, Enterprise, and Education subscribers are excluded from the ad rollout. The company confirmed the test in its own post, which also outlined a set of advertising principles asserting that sponsored content will not influence model responses — what OpenAI calls "answer independence."
The ads appear below ChatGPT responses in a clearly labeled format, separated from organic answers. Targeting is contextual: the model uses the conversation topic, not user identity data, to determine relevance. OpenAI says it does not sell conversation data to advertisers and that users can opt out of ad personalization.
The financial logic driving the decision is not subtle. According to a 2024 report by The Information, OpenAI's own projections showed losses rising as high as $14 billion in 2026. Reuters and CNBC reported in February 2026 that the company expects compute spending to exceed $600 billion through 2030. Those numbers frame the ad product for what it is: a new revenue layer aimed at the bottom of the subscriber pyramid — users who get the service free.
The reversal is also personal for Altman. "We think about advertising as being overlaid or put into a model, and I think in that scenario the answers become slightly worse," he said at an event in 2022. "I think ads are a symptom of a business model failure." The ad rollout announced in early 2026 is a direct contradiction of that position.
On pricing, The Information reported OpenAI is targeting approximately $60 CPM — roughly three times what Meta charges — with a minimum advertiser commitment of $200,000, per reporting from Adweek. That entry barrier effectively limits early participation to large enterprise brands. The premium reflects OpenAI's theory that conversational context is more valuable than a keyword search: a user who has spent several exchanges planning a purchase is further down the decision funnel than someone who typed a query into a search box.
As of October 2025, Sam Altman said ChatGPT had more than 800 million weekly active users. Even a small fraction of that base seeing ads at $60 CPM represents substantial revenue potential — though how many of those users are in the US and logged in as Free or Go subscribers (the populations eligible for this test) is not publicly disclosed.
The structural tension in OpenAI's stated principles is worth watching. The company promises that ads will not influence model outputs. That is easy to assert at the start of a beta and harder to sustain as advertiser pressure mounts, revenue targets grow, and the product matures. Google made similar commitments about search result independence. Whether the conversational format creates more or less pressure on that boundary than a ranked link list is an open question.
For publishers, the picture is uncomfortable. OpenAI has already diverted traffic that used to flow through search; an ad-supported ChatGPT means OpenAI now also captures advertising revenue on top of that diverted attention. Felix Simon at the Reuters Institute noted that news publishers face "yet another indignity in their long disruption" as AI assistants absorb both the audience and the monetization layer.
The more immediate question is simpler: whether the product quality holds. Altman's 2022 concern was that ad integration degrades answers. OpenAI's own principles assert it will not. The test now running for Free and Go users in the US is the first real-world evidence on that question.