OpenAI tests visual ad format in ChatGPT

OpenAI frames advertising as a way to reach a much larger market than subscriptions alone, and it now puts a number on the audience: ChatGPT reaches 1.2 billion people each week. The visual format extends that push by testing images during image generation, while the measurement build-out — integrations with Hightouch, Tealium and LiveRamp, attribution partners such as AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin, and full-funnel partners Fospha, Measured and INCRMNTAL — is aimed at letting advertisers judge results inside the systems they already use.

The early evidence OpenAI cites is third-party and specific: DV Rockerbox's figure of a 15.3% lower attributed cost per acquisition for WeightWatchers against its blended paid-search benchmark, WorkMagic's finding that 67% of Dose's incremental purchases came from net-new customers, and Triple Whale's report that 93% of Portland Leather's visitors from ChatGPT Ads were new. OpenAI also says its incrementality work is still in its early stages, with geo-based experiments alongside Haus, Measured and WorkMagic.

On trust, OpenAI pairs placement guardrails against emotionally vulnerable or sensitive contexts with automated review, human oversight and Negative Phrases for qualifying advertisers, and is developing brand suitability evaluation pilots with DoubleVerify and Integral Ad Science so independent partners can assess how its standards are applied without accessing private user conversations. The open question is whether that structure gives brands enough confidence to shift budgets — the test's reach, the pilots' results and any expansion beyond image generation are the signals to watch.