OpenAI begins U.S. test of ads inside ChatGPT app

OpenAI announced it will begin testing advertisements inside the ChatGPT app for a subset of logged-in users in the United States, placing clearly labeled ad units beneath chatbot answers. The initial experiment is confined to users on the free tier and the lower-cost ChatGPT Go subscription, which costs $8 per month; subscribers to Plus, Pro, Business and Enterprise plans will not see ads.

Ads will appear at the bottom of ChatGPT responses when the company determines there is a relevant sponsored product or service tied to the active conversation. In company mock-ups, the units resemble blocked-off chat sections containing a small image and advertising copy, and they are dismissible. OpenAI said the ads will be visually distinct from the model’s answers, separated from them, and will not influence the chatbot’s outputs. Users can view why an individual ad was shown and have the option to turn off ad personalization entirely.

OpenAI emphasized privacy boundaries: the company said it will "never" sell user data to advertisers, and that advertisers will not gain access to users’ conversation data, age, location or interests for targeting. The rollout will also exclude minors; users under 18 will not be shown ads. The company plans to block advertising related to certain sensitive topics, including politics, health and mental health.

The move marks a notable shift for OpenAI leadership. CEO Sam Altman has previously described advertising in ChatGPT as a "last resort" and said he expected the company to try ads "at some point." Fidji Simo, OpenAI’s chief of applications, framed the test as balanced against the need to preserve trust, writing that users must trust ChatGPT’s responses are "driven by what's objectively useful, never by advertising."

OpenAI characterized the effort as a cautious test rather than a broader monetization overhaul. The company said ads will begin appearing "in the coming weeks" for the targeted U.S. users. The timing follows the global roll-out of ChatGPT Go, which the company introduced after initial launches in markets such as India.

Financial pressures and rising infrastructure costs frame the backdrop for the experiment. Financial documents and projections show OpenAI is committing large sums to data centers and AI chips, with infrastructure spending on the order of $1 trillion to $1.4 trillion over coming years, and forecasts that include billions in operating losses and revenues as the company scales. Only a small fraction of ChatGPT’s large user base pays for subscriptions, complicating efforts to monetize through subscriptions alone.

Analysts and industry observers say advertising could add a meaningful revenue stream but warn it risks eroding trust or driving users to competing chat products. Companies such as Google and Anthropic are already developing rival generative AI assistants, and a shift toward ad-supported chat could become a differentiator.

For now, OpenAI is limiting the experiment and building mechanisms for transparency and control. Whether the test will expand or prompt further changes in product tiers and pricing will likely depend on user reaction, regulatory scrutiny and how effectively the company maintains an editorial firewall between ads and AI-generated answers.