OpenAI begins testing ads in ChatGPT

OpenAI began testing advertising inside ChatGPT, following through on the approach to advertising it had outlined the previous month. The test reached logged-in adult users in the United States on ChatGPT’s Free and Go subscription tiers; Plus, Pro, Business, Enterprise and Edu accounts were excluded. Sponsored placements appear at the bottom of a ChatGPT answer, when the company judges a relevant product or service exists, and are visually separated from and labelled apart from the model’s own response. OpenAI said ads would not influence the content of answers themselves.

Reported initial advertisers included Target, Adobe, Williams-Sonoma and Albertsons. OpenAI said advertisers would not get access to users’ chat logs or personal data, receiving only aggregated performance metrics, and would be charged on a cost-per-view basis tied to impressions rather than clicks.

The launch came days after Anthropic had publicly pledged to keep Claude ad-free, explicitly citing the incentive problems an advertising business model would create for a conversational AI product, and drew criticism — including implicitly from Anthropic’s own marketing — that ads placed directly beneath an AI-generated answer risked blurring the line between the model’s independent output and paid promotion in a way that a search results page, with ads set apart from organic results, traditionally had not.