OpenAI plans visual ads for ChatGPT image generation

OpenAI plans to test a visual advertising format during image generation in ChatGPT with an initial group of advertisers in the United States later in October. It said the ads will be labeled and kept separate from generated images.

The move extends advertising into ChatGPT’s image-generation workflow as OpenAI expands the systems advertisers can use to measure results. It follows OpenAI’s expansion of ChatGPT shopping tools with Albertsons. OpenAI said advertising will not influence ChatGPT’s answers, a product-behavior claim the opened independent coverage did not test.

To track conversions, OpenAI said integrations with Hightouch, Tealium and LiveRamp let advertisers send conversion data from existing systems to ChatGPT Ads. It also listed AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge and Tenjin as partners supporting combinations of its Conversions API, reporting and click-attribution tools.

OpenAI described its incrementality work as early-stage. Incrementality estimates whether an ad caused an outcome that would not otherwise have happened. The company said it is exploring geography-based experiments with Haus, Measured and WorkMagic.

OpenAI also published three partner-supplied performance findings. DV Rockerbox said WeightWatchers’ attributed cost per acquisition was 15.3% below its blended paid-search benchmark. WorkMagic reported that 67% of incremental purchases for Dose came from new customers, while Triple Whale said 93% of Portland Leather visitors from ChatGPT Ads were new. The partners supplied those results, which OpenAI published without sample sizes, campaign periods, uncertainty ranges, methodology or raw data.

OpenAI said ChatGPT reaches 1.2 billion people each week and called it the world’s largest AI-native consumer platform. The opened independent sources did not establish either claim.

For brand controls, OpenAI said it is developing controlled evaluation pilots with DoubleVerify and Integral Ad Science without giving them access to private user conversations. It also said a Negative Phrases control is available to qualifying advertisers with brand-specific placement requirements. The company did not disclose the visual test’s exact October start date, participating advertisers, pricing, eligibility tiers or size.