# ChatGPT Visual Ads: What the Image Generation Test Means Author: Angad Singh Author URL: https://segwise.ai/blog/author/angad-singh Published: 2026-10-12 Category: Strategy & Optimization Category URL: https://segwise.ai/blog/category/strategy-and-optimization Meta Title: ChatGPT Visual Ads: What the Image Generation Test Means Meta Description: ChatGPT visual ads are OpenAI's new image-led ads, tested during image generation. See how they work, how they're measured and what to brief before access. Tags: creative analytics platforms, Creative Generation Tag URLs: creative analytics platforms (https://segwise.ai/blog/tag/creative-analytics-platforms), Creative Generation (https://segwise.ai/blog/tag/creative-generation) URL: https://segwise.ai/blog/chatgpt-visual-ads-image-generation ChatGPT visual ads are image-led ads that OpenAI said on October 5, 2026 it would begin testing later that month, shown to US users on the Free and Go plans while ChatGPT generates an image they asked for. The ads are labeled and kept separate from the image being made. Until now, ChatGPT ads appeared below text responses, per [Search Engine Journal](https://www.searchenginejournal.com/openai-chatgpt-image-generation-visual-ads-test/591981/). In this slot the image leads, so imagery will carry more of the weight. For the wider channel, see our [ChatGPT ads guide](https://segwise.ai/blog/chatgpt-ads-2026-guide). The announcement, titled ["Building advertising for the way people use AI"](https://openai.com/index/new-chatgpt-ads-format-and-measurement/), also widened the list of measurement partners for ChatGPT Ads, shared early advertiser results, and added brand suitability controls. The new unit appears while a person waits for an image to finish. OpenAI has not said how long the ad stays on screen. Segwise is an AI creative analytics and creative generation platform that helps UA and creative teams find their winning creative patterns and turn them into new ads. Facts come from OpenAI and the outlets that covered the launch; where we add our own judgement, we say so. ## ChatGPT visual ads at a glance - **What and where.** Image-led ads shown during image generation to US Free and Go users, with the test starting late October 2026 for an initial advertiser group ( [OpenAI](https://openai.com/index/new-chatgpt-ads-format-and-measurement/)). - **Measurement.** OpenAI names ten click-attribution partners for ChatGPT Ads overall, including the main mobile measurement partners (MMPs), plus early geo-based incrementality experiments ( [OpenAI](https://openai.com/index/new-chatgpt-ads-format-and-measurement/)). - **Not yet published.** Targeting, pricing, creative requirements, reporting and on-screen time for the visual format ( [SEJ](https://www.searchenginejournal.com/openai-chatgpt-image-generation-visual-ads-test/591981/)). - **What to do now.** Brief it like a feed ad, sort your proven images by what transfers, and check your attribution path before asking for access. ## What ChatGPT visual ads look like ChatGPT visual ads are display-style ads built around images rather than text. OpenAI says the format "helps people imagine how products and services could fit into their lives." Its launch post names three kinds of imagery (the examples are ours): 1. Product inspiration, such as a styled spread of ingredients for dinner. 2. Product usage, such as the product in someone's hands or home. 3. Experiences, such as the trip or the meal the product makes possible. The launch mockups use a grocery brand, Heirloom Grocery, to show two layouts on a phone. In both, the brand name sits top left with an "Ad" label top right. One version is a full-width image carousel, one image at a time, with the line "Fresh ideas for dinner." set on the first image. The other is a row of separate product photos that runs off the edge of the screen. Each closes with the headline "Fresh Groceries for Easy Weeknight Dinners" and a "Learn more" button. The ad sits under the progress indicator for the image being generated (62% in the mockup), with the "Ask ChatGPT" input bar still at the bottom. ![OpenAI launch mockup of ChatGPT visual ads for a grocery brand, shown during image generation (image: OpenAI)](https://prod.superblogcdn.com/site_cuid_clo00o2d0644641vqp8vh8w6cd/images/chatgpt-visual-ads-what-the-image-generation-test--image-1-1791786199225-compressed.jpg) _Mockup: OpenAI, from its October 5, 2026 announcement._ ### How visual ads differ from the ads ChatGPT already runs Existing ChatGPT ads New visual ads (test) Where they appear Below text responses During image generation What the user is doing Asking a question Waiting on an image, a tool-style task Availability Live; expanded to India in August ( [TechCrunch](https://techcrunch.com/2026/10/05/openai-launches-visual-ads-that-appear-alongside-image-generation-results/)) US test with an initial advertiser group What carries the ad (our read) Headline and relevance to the question The image SEJ's Brooke Osmundson sums up the availability row: most advertisers will not have access yet, which makes this "more of a format to watch than something advertisers need to plan around today." ## Why OpenAI put ads in image generation first Asad Awan, who leads monetization product and engineering at OpenAI, described this placement on the [Mobile Dev Memo podcast](https://mdm.co/podcast-openais-advertising-ambitions-with-asad-awan/) on September 30, five days before the launch. His point is that some ChatGPT use is tool-style. You are not asking a question. You are making something. "I am waiting for the image to get generated, that is a really good place for us to experiment," he said. In those moments, "I am not in the contextual mode anyway, I am using the product like a tool not to answer a specific thing." Eric Seufert, who hosted the episode, calls the new unit "a display ad" in his [write-up of the launch](https://mdm.co/openai-introduces-visual-ads-to-chatgpt/), shown "in a context that likely includes little to no commercial intent." Pointing back to his essay [The privileged position of chatbot advertising](https://mdm.co/the-privileged-position-of-chatbot-advertising/), he argues chatbot ads enjoy an "either/or benefit." If a conversation has a commercial angle, the ad can be anchored to it. If it does not, the ad can be anchored to the conversion data advertisers send through their Conversions API (CAPI) setups or pixels. An ad card under a shopping question behaves like search. A visual ad shown while someone waits for a birthday card illustration behaves like a feed ad, where the image does the work. Seufert makes the same comparison: a chatbot ad "can look exactly like an ad in a feed." > **Brief it like a feed ad** Judge creative for this slot on whether it stops someone busy with something else, not on how well the copy matches a query. ## What ChatGPT visual ads mean for UA and creative teams The work you can do before access opens is mostly on the creative side. ### Sort your existing image signals into "transfers" and "doesn't" You already have evidence on which images work with a cold audience. Not all of it carries over to this slot. Likely to transfer Unlikely to transfer Static and carousel winners from Meta, and TikTok photo or carousel ads Playable mechanics and end cards Product shots, colour palettes and compositions that win with cold audiences Gameplay or UI screenshots that need context to make sense Lifestyle images that show the outcome or the experience Offer-led text overlays written for search intent Images that read at a glance with no sound Hooks that rely on the first seconds of a video The split follows from the format: a static image or carousel, seen by someone who is busy with something else. ![Comparison of image signals likely to transfer to ChatGPT visual ads versus those unlikely to transfer](https://prod.superblogcdn.com/site_cuid_clo00o2d0644641vqp8vh8w6cd/images/chatgpt-visual-ads-what-the-image-generation-test--image-2-1791786200527-compressed.jpg) ### Cover all three of OpenAI's image themes Count what you have in each theme before the format opens up, and brief whichever theme is thinnest first. Here is how the themes map for the kinds of advertisers Segwise works with: - **Mobile games.** Gameplay UI rarely reads at a glance, so it is a weak fit. Key art, characters and world imagery fit the experience theme, and that is where a game should start. - **Subscription and consumer apps.** Show the outcome, not the app screen: the finished meal plan, the calmer morning, the trip that got booked. That covers usage and experience. - **DTC and ecommerce.** Styled product spreads cover inspiration, the product in a real home covers usage, and the moment it enables covers experience. Most catalogs can fill all three. ### Use your tagged data to pick the first test set [Segwise's Creative Tagging Agent](https://segwise.ai/features/creative-tagging) tags the colours, products, characters, on-screen text, emotional tone and visual styles in every image and video you run, and maps each tag to the metrics you track, such as ROAS, CPA, IPM and CTR, across 15+ ad networks and your MMP. Filter that to the "likely to transfer" column and you have a data-backed shortlist for a ChatGPT test. Segwise does not connect to ChatGPT Ads today. It is where you find your proven visual patterns before the test, not a ChatGPT dashboard. Know which images win before ChatGPT opens up See which colours, products and visual styles drive ROAS across your current networks [See Segwise creative tagging](https://segwise.ai/features/creative-tagging) [Start Free Trial](https://ua.segwise.ai) ## How ChatGPT Ads are measured OpenAI's partner list covers ChatGPT Ads in general. It has said nothing specific about measuring the visual format. OpenAI supports both in-house and partner-led measurement, and named partners for four jobs ( [OpenAI](https://openai.com/index/new-chatgpt-ads-format-and-measurement/)): Job Partners named by OpenAI Send conversion data in Hightouch, Tealium, LiveRamp Click attribution, reporting and Conversions API (web and app) AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, Tenjin Full-funnel and advanced measurement Fospha, Measured, INCRMNTAL Exploring geo-based incrementality experiments (early stage) Haus, Measured, WorkMagic For app marketers, the attribution row is the one that matters. AppsFlyer, Adjust, Branch, Singular and Kochava are all on it, so ChatGPT ad clicks can in principle land in the same MMP you use for Meta, Google, TikTok and Axon (AppLovin). OpenAI has not mentioned view-through attribution or SKAdNetwork, so do not assume either. ### The early results, and how to read them OpenAI shared three partner-reported results: - **WeightWatchers:** attributed CPA 15.3% lower than its blended paid search benchmark, per DV Rockerbox. - **Dose:** statistically significant lift, with 67% of incremental purchases from net-new customers, per WorkMagic. - **Portland Leather:** 93% of visitors from ChatGPT ads were new, per Triple Whale. These measure three different things: attributed CPA, incremental lift and new-visitor share. SEJ makes the same point and warns against treating them as general benchmarks. One more caveat: OpenAI does not say which placement produced them, but the visual test had not started, so they reflect the existing below-answer ads and say nothing yet about the visual slot. When it opens, the new-customer share is the number to watch. ## Brand safety controls for the new format An ad next to something a person is creating raises an obvious question: what if the image is about something sensitive? OpenAI lists three controls ( [OpenAI](https://openai.com/index/new-chatgpt-ads-format-and-measurement/)): - **Placement guardrails** judge whether a conversation suits ads at all, to keep ads out of emotionally vulnerable, sensitive or unsuitable contexts. OpenAI also uses automated review, human oversight and monitoring to enforce its ad policies. - **Negative Phrases** let qualifying advertisers keep ads away from topics that clash with their own brand policies. - **Brand suitability pilots** with DoubleVerify and Integral Ad Science let independent partners test those safeguards without access to private conversations. On the Mobile Dev Memo podcast, Awan used the example of photographing his kid's bruise to ask ChatGPT about it, to explain why people must be able to trust ChatGPT with personal information. Later in the conversation he agreed personal topics like that should not be turned into a commercial opportunity. If your category sits near topics like that, ask about Negative Phrases before you ask about reach. ## What OpenAI has not published about ChatGPT visual ads SEJ lists the gaps in OpenAI's announcement, including that there is no timeline for a broader rollout: - **Pricing.** No CPMs or bid model for the visual format. - **Targeting.** No detail on how visual ads are matched to people during image generation. - **Creative specs.** No image sizes, aspect ratios, copy limits or carousel rules. - **Reporting.** No format-level reporting described. - **On-screen time.** No detail on how long the ad shows while an image generates. - **Rollout.** A global rollout is planned, per TechCrunch, but neither it nor SEJ gives a date, and the test has no published end. ## Who should push for early access to ChatGPT visual ads OpenAI has not said how advertisers join the visual test, beyond inviting businesses to sign up at ads.openai.com, per [its announcement](https://openai.com/index/new-chatgpt-ads-format-and-measurement/). It has also published no spend minimum for the test. This is how we would decide whether to apply. ![Three checks before asking for ChatGPT visual ads access: attribution ready, visual product, geo holdout](https://prod.superblogcdn.com/site_cuid_clo00o2d0644641vqp8vh8w6cd/images/chatgpt-visual-ads-what-the-image-generation-test--image-3-1791786201984-compressed.jpg) Ask now if all of these are true: - Your MMP or attribution tool is on OpenAI's list (AppsFlyer, Adjust, Branch, Singular, Kochava, or another named partner). - Your product is easy to show in one image. Food, home, fashion, travel, games with strong key art, and apps with a visible outcome fit OpenAI's themes. A B2B integration or a feature-heavy utility app does not. - You can run a geo holdout of two to four weeks with an incrementality partner (OpenAI names Haus, Measured and WorkMagic for early geo tests), and your paid social already drives roughly 100 or more conversions per region per week, so a lift can show up. OpenAI sets no such bar; that is our rule of thumb for a first geo test. - You have at least five or six images that fit the "likely to transfer" column, so the test is not riding on one creative. Wait if your ads would often sit next to sensitive conversations, until you know how Negative Phrases work in practice. Note that this is about the topic, not the vertical: WeightWatchers and Dose, both health-adjacent brands, are among OpenAI's showcase results. A clean measurement path and an image set sorted by theme also help on the feed placements you already run. When you want that image evidence from the ads you run today, Segwise helps your team see which visual elements drive results across your networks and MMP, then turn them into new on-brand statics and videos. ## Frequently asked questions ### What are ChatGPT visual ads? ChatGPT visual ads are image-led ads that OpenAI announced on October 5, 2026. They show product inspiration, product usage, or the experiences a product makes possible. The first test places them in ChatGPT during image generation, clearly labeled and separate from the image being created ( [OpenAI](https://openai.com/index/new-chatgpt-ads-format-and-measurement/)). ### What do ChatGPT visual ads mean for app marketers? They add a feed-style visual placement inside ChatGPT, which reaches 1.2 billion people weekly per [OpenAI](https://openai.com/index/new-chatgpt-ads-format-and-measurement/), though the first test covers only US Free and Go users. OpenAI [names](https://openai.com/index/new-chatgpt-ads-format-and-measurement/) the major MMPs as attribution partners for ChatGPT Ads generally, but has not detailed measurement for the visual format itself. ### What kind of images work best for ChatGPT visual ads? OpenAI says the format is built for images of product inspiration, product usage, or the experiences a product makes possible ( [OpenAI](https://openai.com/index/new-chatgpt-ads-format-and-measurement/)). It has not published creative specs. Because the ad is seen by someone waiting on their own image, our view is that images which read at a glance with no sound or context fit best: lifestyle and outcome shots, styled product spreads, and game key art, rather than UI screenshots or offer-heavy text. A creative analytics tool such as Segwise can show which of those image traits already win with cold audiences on your other networks. ### What's the difference between ChatGPT visual ads and regular ChatGPT ads? Regular ChatGPT ads appear below text responses, according to [Search Engine Journal](https://www.searchenginejournal.com/openai-chatgpt-image-generation-visual-ads-test/591981/). Visual ads appear while ChatGPT generates an image and lean on imagery rather than copy. That makes them closer to feed display ads than search ads, a point Eric Seufert makes in his [Mobile Dev Memo analysis](https://mdm.co/openai-introduces-visual-ads-to-chatgpt/). ### Is it worth moving budget to ChatGPT image ads yet? Not for most teams. Access is limited to an initial US advertiser group, and pricing, targeting and specs are unpublished. [SEJ](https://www.searchenginejournal.com/openai-chatgpt-image-generation-visual-ads-test/591981/) argues a new format alone does not justify shifting budget. Set up measurement and sort your image library now so you can move quickly when access opens. ## FAQs Q: What are ChatGPT visual ads? A: ChatGPT visual ads are image-led ads that OpenAI announced on October 5, 2026. They show product inspiration, product usage, or the experiences a product makes possible. The first test places them in ChatGPT during image generation, clearly labeled and separate from the image being created (OpenAI). Q: What do ChatGPT visual ads mean for app marketers? A: They add a feed-style visual placement inside ChatGPT, which reaches 1.2 billion people weekly per OpenAI, though the first test covers only US Free and Go users. OpenAI names the major MMPs as attribution partners for ChatGPT Ads generally, but has not detailed measurement for the visual format itself. Q: What kind of images work best for ChatGPT visual ads? A: OpenAI says the format is built for images of product inspiration, product usage, or the experiences a product makes possible (OpenAI). It has not published creative specs. Because the ad is seen by someone waiting on their own image, our view is that images which read at a glance with no sound or context fit best: lifestyle and outcome shots, styled product spreads, and game key art, rather than UI screenshots or offer-heavy text. A creative analytics tool such as Segwise can show which of those image traits already win with cold audiences on your other networks. Q: What's the difference between ChatGPT visual ads and regular ChatGPT ads? A: Regular ChatGPT ads appear below text responses, according to Search Engine Journal. Visual ads appear while ChatGPT generates an image and lean on imagery rather than copy. That makes them closer to feed display ads than search ads, a point Eric Seufert makes in his Mobile Dev Memo analysis. Q: Is it worth moving budget to ChatGPT image ads yet? A: Not for most teams. Access is limited to an initial US advertiser group, and pricing, targeting and specs are unpublished. SEJ argues a new format alone does not justify shifting budget. Set up measurement and sort your image library now so you can move quickly when access opens. --- This blog is powered by Superblog. Visit https://superblog.ai to know more. ---