OpenAI publishes its approach to advertising in ChatGPT

OpenAI set out, for the first time, how it intended to run advertising inside ChatGPT — publishing a set of principles alongside the announcement of its low-cost ChatGPT Go tier, ahead of testing due to begin in the US in the following weeks. The post committed to keeping ads separate from — and non-influential on — the underlying answers: “Ads do not influence the answers ChatGPT gives you,” the company wrote, describing answers as optimised for what’s most helpful regardless of advertiser relationships, with ads “always separate and clearly labeled.” It also committed to keeping user conversations private from advertisers, saying data would never be sold.

Only logged-in adult users on the free tier and subscribers to the $8-a-month Go plan would see ads when testing began; Plus, Pro, Business and Enterprise subscribers, and any account identified as under 18, would not. Ads were to appear at the bottom of a response, only when a sponsored product or service was relevant to the conversation, and would be withheld near sensitive or regulated topics such as politics, health or mental health. A worked example in the announcement showed a travel query about Santa Fe surfaced with a cottage-rental advertiser, including an option to open a direct chat with the advertiser’s own bot.

The announcement marked OpenAI’s clearest public signal that it would follow the standard consumer-internet path toward advertising revenue to subsidise a free product, after two years of running ChatGPT’s free tier without ads. Commentators, including independent developer Simon Willison, noted that OpenAI’s assurance addressed direct manipulation of answers but left open whether the optional advertiser-chat feature could itself become a vector for commercial influence over what ChatGPT told users. The plan came to fruition in February, when OpenAI began testing ads with a first cohort of US users.