- Visual Ads: OpenAI will test visual ads during image generation in its ChatGPT AI chatbot in the U.S. later in October.
- Conversion Data: Expanded tools connect advertisers’ purchase and signup data to campaign reports and help guide which ads OpenAI delivers.
- Free Option: Ads-Free removes ads on the Free plan but reduces limits and access to tools such as image generation.
OpenAI plans to show visual ads during image generation in ChatGPT, its AI chatbot, adding product promotions while users create pictures. The U.S. test is due later in October, with ads labeled and kept separate from the image being made.
Alongside the planned visual test, OpenAI is expanding advertiser measurement tools and developing checks on where ads appear. OpenAI puts ChatGPT’s audience at 1.2 billion people each week and presents advertising as a way to serve more users than subscriptions alone.
Visual Ads During Image Generation
The proposed visuals are intended to show products in use, offer inspiration or depict experiences associated with a service. Existing ChatGPT ads already contain an image alongside an advertiser’s name, headline and description. The new test adds a promotional encounter during picture creation.
ChatGPT’s original U.S. ad test began on February 9 for Free and Go users, with ads separate from the assistant’s response. Under the broader service’s ad eligibility and controls, Free and the paid Go plan may carry ads; Plus, Pro, Business, Enterprise and Edu remain ad-free.
Eligible Free users can choose an existing Ads-Free experience, but it lowers usage limits and restricts access to tools such as image generation. Turning off ad personalization instead keeps ads based on the current chat. OpenAI has not specified the visual pilot’s exact user subset or start day.
In May 2026, Google described Sponsored entries within AI Mode product recommendations. AI Mode is its AI-assisted search service; that format puts advertiser products inside a recommendation list. OpenAI’s proposed visual placement keeps advertising outside the image ChatGPT creates.
Customer Actions Feed Reports and Ad Delivery
OpenAI was already building pixel and API tracking tools in May. The October expansion adds connections to advertisers’ existing systems through Hightouch, Tealium and LiveRamp, letting businesses send purchase and other customer-action data into ChatGPT Ads.
A conversion is a recorded action such as a purchase or signup. The Measurement Pixel is tracking software on an advertiser’s website; the Conversions API lets its server send events directly to OpenAI. The event data gives advertisers feedback on what people do after encountering a campaign.
Attribution reports assign campaign credit for customer actions. Advertisers can choose how long after a click a purchase receives credit and add a one-day view-through window for conversions following an impression, a recorded ad showing, without a click. OpenAI’s advanced conversion optimization separately learns from eligible click-through and view-through conversions to guide future ad delivery. Changing a reporting window changes credited results, rather than how the campaign optimizes.
Credited Actions and Additional Sales
Incrementality asks how many additional purchases the advertising caused. OpenAI’s work on geographic experiments with Haus, Measured and WorkMagic remains in its early stages.
For existing ChatGPT campaigns, OpenAI cites measurement provider DV Rockerbox’s finding that the advertiser WeightWatchers’ attributed cost per acquisition was 15.3% below its blended paid-search benchmark, a combined reference for paid advertising in search engines. The metric expresses advertising spending per action credited as an acquisition. OpenAI does not disclose which actions were counted, the campaign mix or the measurement period.
In Digiday’s October 1 report, advertising agency Accuracast’s CEO Farhad Divecha described a U.K. campaign that received completed forms but still showed zero conversions in ChatGPT’s reports two weeks later. The agency also saw ChatGPT report around 100 clicks while its own tracking recorded 20.
Some advertisers withheld conversion tracking over their legal teams’ concerns about customer data and OpenAI’s terms. An agency said clients had longer familiarity with Google and Meta’s agreements. Businesses withholding those records give OpenAI less feedback through conversion tracking to inform reports and later ad delivery.
Advertising executives at AdRoll and Jellyfish said they had not seen these click-count discrepancies. Jellyfish senior paid media director Victor Batista called the shift toward cost-per-acquisition optimization pivotal. Using acquisition feedback rather than clicks alone gives ad delivery a customer outcome to pursue; Batista expects confidence to improve as more partner data is analyzed.
Ad Suitability Depends on the Conversation
OpenAI says its placement safeguards assess whether conversations are suitable for advertising, seeking to keep ads away from emotionally vulnerable or sensitive contexts. Qualifying advertisers can also set narrow, brand-specific placement restrictions with Negative Phrases.
Verification firms DoubleVerify and Integral Ad Science are developing controlled evaluations of those safeguards. DoubleVerify CEO Mark Zagorski identifies the conversation, rather than a public page or video, as the context surrounding a ChatGPT ad. OpenAI says the firms will assess its standards without access to private user conversations.
Unpredictable Ad Delivery
Accuracast told Digiday that one campaign served thousands of impressions one week and fewer than 100 the next. Divecha described unpredictable inventory, the available opportunities to show ads, as the agency’s biggest problem.
