OpenAI bets on ads as rivals sell an ad-free future

232

OpenAI, Anthropic and Perplexity are turning advertising into a test of trust, not just a revenue model. The split is accelerating as AI companies seek durable revenue to fund expensive infrastructure. OpenAI began testing clearly labeled ads below ChatGPT responses for logged-in adult US users on Free and Go plans in February, while Anthropic committed to keeping Claude ad-free and Perplexity stepped back from advertising. The divide matters because conversational AI increasingly sits between consumers and purchase decisions.

OpenAI says, “Ads do not influence the answers ChatGPT gives you,” and says advertisers cannot access conversations. Its case is scale: OpenAI research reported more than 700 million weekly active consumer users by the end of July 2025. Even limited inventory across that audience could create a high-intent alternative to search advertising.

Anthropic is making the opposite bet. “There are many good places for advertising. A conversation with Claude is not one of them,” the company said, arguing that open-ended chats often contain personal or sensitive context. Perplexity, which began testing sponsored questions in 2024, has abandoned advertising for now over concerns that users would doubt the neutrality of answers.

For marketers, the opportunity is narrower than the hype. AI advertising will reward structured product data, transparent sponsorship and measurable downstream actions, not banner-style creative transplanted into a chat window. Research on AI agents suggests they respond more consistently to price, availability and specifications than to visual persuasion.

The strategic question is not which chatbot offers the largest audience, but which can commercialize intent without weakening answer credibility. Brands should test cautiously, separate paid visibility from organic AI optimization, and track trust signals alongside clicks. In conversational AI, a cheap conversion can become expensive if users start questioning every recommendation.