OpenAI is getting ready to widen ChatGPT ad sales next month, and it’s already doing some course-correcting after a first pilot that left early buyers squinting at the dashboard. Agencies working with initial advertisers say the buying process felt decidedly “2009”—phone calls, emails, and spreadsheets—and the reporting hasn’t gone much further than views and clicks. Result: two agency execs said they still can’t point to measurable business outcomes for clients.
The early economics were anything but beta-priced: OpenAI charged $60 CPM (a rate typically reserved for truly scarce, high-heat inventory) and required at least a $200,000 commitment per advertiser. Yet delivery has been slow. One exec said ads weren’t shown frequently enough to generate meaningful learnings, and another said clients had only spent ~15%–20% of committed budgets even though the pilot is more than halfway done. OpenAI says results vary, but it’s seeing week-over-week progress—and it’s now increasing the number of users who see ads, with plans to show ads to all free U.S. users (and the lowest-cost tier) in the coming weeks.
To make this easier to buy—and maybe easier to defend in a QBR—OpenAI is considering either partnerships with ad tech firms or its own ad management system; it’s reportedly testing a self-serve ad manager with some partners (no public timeline yet). It also announced a Criteo integration that looks primarily like a purchasing interface for now, with Criteo pitching $50,000–$100,000 commitments; note that pilot spend won’t count toward that. Targeting help may be limited in the near term because OpenAI is sharing only basic, aggregated data (impressions, clicks, spend), and OpenAI says it’s not pursuing outside-run programmatic auctions right now—so for smaller brands that need tight attribution, agencies are already hinting: proceed only if you’ve got “experimental budget” money and a tolerance for ambiguity.
Read more at The Information.