OpenAI's measurement gaps keep ChatGPT ad budgets at test level

Seven months after OpenAI started selling ads in ChatGPT, measurement remains the sticking point in nearly every conversation marketers have about the company as an ad seller. The result: budgets are staying at test level, and there is little sign yet of a shift.

The frustration has two roots, according to marketers Digiday has spoken with. First, a sense that OpenAI hasn't built much beyond the basics in its ad product. Second, marketers' own reluctance to share their data with the company — a trust problem OpenAI has to solve on its end before spend scales.

On the product side, agencies running campaigns describe gaps that would be unusual on Google or Meta. Accuracast ran a U.K. lead-generation campaign on ChatGPT alone. The form fills came in. Two weeks later, the platform still showed zero conversions.

Accuracast's group CEO Farhad Divecha pointed out the contrast with the duopoly: on Google and Meta, the problem tends to run the opposite way, with the platforms claiming more conversions than they actually drove. ChatGPT, in Accuracast's experience, underreports.

The discrepancies don't stop at conversions. The same agency found ChatGPT reporting around 100 clicks where its own tracking picked up 20. OpenAI's support team, at the time, couldn't explain why.

It's not a universal problem, though. Executives from AdRoll, Jellyfish and another independent U.S. ad agency said they hadn't seen click discrepancies in their campaigns. That split matters for practitioners: before writing off ChatGPT ads on reporting grounds, advertisers should run their own parallel tracking and compare, because experiences vary.

For Web Guide Partner, the issue is timing rather than accuracy. Where conversions do register on ChatGPT, the agency sees them arrive 24 to 36 hours late. On Google, the same data lands within a few hours. For any campaign that requires fast optimization decisions — dayparting, budget reallocation, creative swaps — a lag of more than a day effectively blinds the buyer during the window when adjustments matter most.

The reporting is also thin on context. ChatGPT's ad reporting currently carries no demographic data and no prompt data, leaving advertisers without the audience dimensions they use routinely on Meta and Google to refine targeting and creative.

For practitioners weighing the channel, the budget implication is straightforward. Until conversion tracking, click reporting and latency match the standards of established platforms — and until OpenAI offers the demographic and contextual reporting marketers expect — ChatGPT ads work as an experiment, not as a line item that can be scaled or held to standard performance benchmarks. Agencies allocating staff time should plan for manual verification: own tracking, tolerance for delayed data, and no assumption that platform-reported numbers will reconcile.

The pattern is familiar from other ad platforms in their first year. But OpenAI is entering a market where Google and Meta set the reporting baseline, and marketers measure every new entrant against it. Whether OpenAI closes these measurement gaps — and earns enough trust for advertisers to share their data — will decide whether ChatGPT ad budgets move beyond test level in the months ahead.