On Monday, 5 October, OpenAI rolled out a visual advertising format for ChatGPT alongside an expanded suite of measurement tools made available to advertisers. Visually, the updated advertisements draw on product inspiration, real-world usage, and the broader experiences a product can enable, while the revised measurement framework connects ChatGPT Ads to the analytics platforms advertisers already have in place.
Image generation inside ChatGPT serves as the initial testing ground for the new format, with rollout limited to accounts on the Free and Go tiers. According to OpenAI's announcement, any advertisements will appear with clear labels and in a distinct space from generated imagery, with the company stating that paid placements carry no bearing on the responses ChatGPT produces.
A select group of advertisers in the U.S. will take part in a pilot phase set to get underway later in October. Advertiser conversion data housed in first-party systems can flow into ChatGPT Ads through tie-ins with three data platforms — Hightouch, Tealium, and LiveRamp.
Attribution partnerships for OpenAI now extend across a roster of ten mobile and web measurement platforms — AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, and Tenjin — while full-funnel measurement capabilities are handled through separate arrangements with Fospha, Measured, and INCRMNTAL. The company characterises its incrementality work as being at an early phase, and is conducting geography-based tests in partnership with Haus, Measured, and WorkMagic to determine whether ChatGPT Ads drive conversions that would not otherwise have occurred — rather than merely capturing attribution for purchases already in motion.
Pilot programmes pairing the platform with both DoubleVerify and Integral Ad Science on brand suitability are currently taking shape. Under a controlled testing arrangement, those companies are able to evaluate how OpenAI's brand safety standards get interpreted and applied — without any exposure to private user conversations.
Paired with the rollout were a handful of figures from initial partners. Measurement data from DV Rockerbox placed WeightWatchers' attributed acquisition cost through ChatGPT Ads at 15.3% under its blended paid-search baseline, while a separate Triple Whale analysis of Portland Leather's traffic revealed that nine in ten visitors arriving via that same channel had no prior purchase history with the brand. Those numbers traced back to individual partners rather than any platform-wide testing programme — a distinction the publication took care to draw out explicitly.
Advertisers running live campaigns on the platform began receiving weekly email digests on Friday, 16 October, laying out a side-by-side comparison of that week's performance against the one before, while the Ads Manager account menu was updated to let users flag technical problems or put forward ideas for improvement. Within Ads Manager, a reworked Conversions page brings conversion activity, event trends, optimisation status, and diagnostic warnings into a single unified view.
Broader ad card formats and full-width image layouts — spanning both desktop and mobile surfaces — are among the display variations currently under evaluation at OpenAI. Participation in those trials draws on advertisers' pre-existing campaign materials and creative assets, meaning no additional preparation is needed from those involved.
Plans are in place to trial a distinct multi-image format aimed at image generation, with a rollout beginning among a limited group of advertisers based in the United States. OpenAI puts ChatGPT's weekly user base at 1.2 billion, with the company adding that revenue from advertising is intended to expand access to its AI tools.
The advertising pilot for the U.S. market got underway on 9 February, according to a news outlet's account. No performance data or wider deployment schedule for the new visual formats has been put into the public domain by OpenAI, a news outlet reported.
