OpenAI Will Test Visual Ads During ChatGPT Image Generation

ChatGPT users could see a clearly labeled visual advertisement while waiting for an image to generate. The ad would appear separately from the requested image, placing advertising inside a generative workflow without making sponsored content part of the output.

OpenAI described the placement in its visual-ad announcement on October 5, 2026. Testing is scheduled to begin later in October in the United States with an initial group of advertisers. The format remains an announced experiment, not a generally available ChatGPT feature.

The announcement also includes conversion-data integrations and plans for independent brand-suitability evaluation pilots. Each deserves separate scrutiny. The placement affects what users encounter; the measurement tools affect how advertisers assess results. The proposed evaluations could test advertising safeguards without exposing private conversations.

The Ad Appears During Generation, Not Inside the Image

OpenAI says the visual format can show product inspiration, product usage, or experiences associated with a product. Its initial placement will be during image generation in ChatGPT.

According to the company, ads will carry clear labels, remain separate from the image being generated, and have no influence over ChatGPT’s answers. The announcement does not describe inserting a sponsor’s product into the generated artwork.

People use image generation to produce something specific. A sponsored visual appearing nearby could be mistaken for an intermediate result, a recommendation from ChatGPT, or part of the requested output. The label needs to communicate both that someone paid for the placement and that it is distinct from the generation.

The format also puts advertising into the waiting period within a creative task, instead of appearing only around a text response. Whether that feels unobtrusive or distracting will depend on interface details the announcement does not fully establish.

OpenAI has not specified the exact layout, how frequently eligible users might encounter the format, or whether they can dismiss it. Its commitment to separation gives users a starting point for evaluating the test, but leaves those implementation questions open.

A Limited U.S. Test Is Not a General Rollout

The announced scope is narrow: the United States, an initial advertiser group, and a start sometime later in October. OpenAI has not named the initial advertisers, given an exact start date, or disclosed the volume of users eligible for the experiment. It has also provided no schedule for expanding the format beyond the initial test.

Without those details, there is no reliable way to predict who will encounter a visual ad.

Plan eligibility is a separate question. OpenAI’s supplied ChatGPT advertising help documentation says ads may appear for Free and Go users, while Plus, Pro, Business, Enterprise, and Edu accounts do not have ads. That describes the existing advertising framework. It does not mean every U.S. Free or Go user will enter the visual-ad test, or that the format will appear every time an eligible user generates an image.

For marketers, OpenAI’s advertiser portal provides a route to sign up for its advertising offering. Signing up does not guarantee access to the initial visual-format experiment, according to the scope established in the announcement.

OpenAI is testing an additional placement within ChatGPT. Image-generation ads have not been announced as available to everyone.

Advertisers Gain New Conversion-Measurement Connections

The measurement changes extend beyond the visual format. OpenAI announced integrations with Hightouch, Tealium, and LiveRamp that let advertisers send conversion data from their existing systems to ChatGPT Ads.

Conversion data records actions advertisers care about, such as purchases, leads, installs, and sign-ups. OpenAI says its measurement offering supports signals from its pixel and Conversions API, alongside partner-led attribution, incrementality, and brand measurement.

Connecting existing systems could reduce the work required to include ChatGPT campaigns in an advertiser’s reporting. Businesses still need to understand which events and fields they send, how those signals are used, and what consent requirements apply to their implementation.

OpenAI names attribution partners including AppsFlyer, Triple Whale, Adjust, and DV Rockerbox. It also describes advanced measurement arrangements with providers such as Fospha, Measured, and INCRMNTAL, plus early-stage geo-based incrementality work with Haus, Measured, and WorkMagic.

These services address different questions. Attribution assigns credit for a conversion according to a reporting method. Incrementality examines whether advertising produced additional outcomes that would otherwise not have happened. Brand measurement investigates effects such as changes in brand perceptions.

Advertisers can also select a click-through attribution window and add a one-day view-through window, OpenAI says. A view-through conversion credits an outcome after an ad impression even when the person did not click.

According to the company, those reporting settings do not change campaign optimization. Optimization can draw on eligible click-through and view-through conversions for the selected campaign goal, so advertisers should distinguish what their reports display from which signals the delivery system uses.

Attributed Sales Are Not Automatically Incremental Sales

OpenAI’s announcement includes early advertiser-performance figures. These are company-published case studies, not independently audited evidence of platform-wide effectiveness.

For example, OpenAI cites DV Rockerbox’s reporting that WeightWatchers’ attributed cost per acquisition on ChatGPT Ads was 15.3% lower than its blended paid-search benchmark. The comparison involves one advertiser and an attribution framework; it does not establish that ChatGPT Ads will outperform paid search for other businesses.

View-through measurement makes the distinction especially important. Someone might see an ad and later purchase from the advertiser, but the sequence alone does not prove the ad caused the purchase. Existing customer interest, another campaign, or an unrelated recommendation could also contribute.

An attribution system can legitimately count that purchase under its rules while leaving the causal question unanswered. Incrementality experiments investigate that question by comparing outcomes under conditions with different advertising exposure.

OpenAI describes its geo-based incrementality work as early-stage. The announcement does not provide an independent audit establishing that the cited results generalize across advertisers, placements, or campaign objectives.

Advertisers should also avoid assuming that results from existing ChatGPT campaigns predict performance during image generation, where the placement occurs within a different task. A useful evaluation would examine the visual format itself, disclose the attribution settings, and separate reported conversions from experimentally measured additional sales.

Brand-Suitability Pilots Will Test Safeguards, Not Certify Them Yet

OpenAI says it is developing brand-suitability evaluation pilots with DoubleVerify and Integral Ad Science. Under the proposed arrangements, the partners would evaluate its safeguards in controlled environments without receiving access to private user conversations. The pilots are proposed independent evaluations, not completed certifications of the advertising system.

Brand suitability concerns whether an advertising placement fits a company’s standards. In a conversational product, that assessment involves the surrounding interaction, rather than simply the subject of a webpage or video.

OpenAI says its placement guardrails assess whether a conversation is appropriate for advertising, aiming to exclude emotionally vulnerable, sensitive, or otherwise unsuitable contexts. It describes automated review, human oversight, and ongoing monitoring as parts of those safeguards.

Controlled testing could help examine how consistently the system handles difficult scenarios while avoiding disclosure of real conversations. The value of that evidence will depend on the scenarios tested, the criteria used, the failures identified, and how much of the findings becomes public.

OpenAI’s broader advertising principles commit to answer independence, conversation privacy, and user control. Those remain company commitments. Announcing external evaluation partners does not, by itself, verify that all three hold throughout deployment.

There are also two separate privacy questions. The promise that suitability partners will not access private conversations concerns the evaluation arrangement. Advertisers sending conversion events through measurement integrations involves a different data flow. One assurance does not fully explain the other.

The most useful next evidence would be a documented user-facing implementation and published evaluation methods and findings. Clear separation between the ad and generated image is necessary. A credible test also needs to show that placement safeguards work and that commercial measurement remains distinct from the substance of ChatGPT’s responses.