On October 1, 2026 OpenAI launched two shopping features worldwide in ChatGPT: Try On, which previews clothing and accessories on a photo the user uploads, and Favorites, which saves products into the ChatGPT Library. Try On appears as a button on eligible listings and runs on ChatGPT Images 2.5. Reference photos are stored for later try-ons until the user deletes them. Retail, marketplace, and performance-marketing teams should check whether their product feeds surface in ChatGPT shopping results and how returns and image-rights language cover a synthetic try-on.
OpenAI’s October 1 shopping update puts a visual try-on and a save-for-later folder inside ChatGPT, on mobile and the web, rather than only inside a retailer’s own site. Ecommerce, marketplace, and performance marketers who already buy or optimize for AI shopping surfaces need a feed and creative check this week, not a brand campaign about virtual fitting rooms.
What changed?
ChatGPT added a Try On button for clothing and accessories and a Favorites save into Library. Users can also upload a product image and ask for a try-on. The image step uses ChatGPT Images 2.5.
When?
Announced October 1, 2026, as a global launch on mobile and web. Images 2.5 itself was introduced earlier, on September 8, 2026.
Who is affected?
Apparel and accessory brands, marketplaces feeding ChatGPT shopping results, and performance teams measuring assisted conversion outside their own site.
What to do this week?
Test whether hero SKUs show a Try On button. Confirm product titles, images, and availability in the feed. Brief service teams that a generated preview is not a fit guarantee.
What did OpenAI ship on October 1?
TechCrunch reported that OpenAI announced a global launch of two shopping features on October 1, 2026. Try On lets a user upload a selfie or a full-body photo and see a clothing item or accessory on that image. The control appears as a Try On button in ChatGPT shopping results. A user can also upload a screenshot or photo of an item and ask ChatGPT to try it on.
Favorites lets a user save a product into a Library in the app and return to it later. OpenAI said saved products sit alongside try-on images. Reporting on the release notes adds that users can organize finds in folders in that Library. The same update mentioned a multi-page document scanner rolling out in the iOS camera. That scanner is a capture tool, not a shopping feature, and marketing teams can ignore it unless field teams use ChatGPT to scan sell sheets.
OpenAI said the shopping visuals use ChatGPT Images 2.5, which it launched on September 8 and describes as better at lighting, texture, instruction following, and keeping a person recognizable from a reference photo. The try-on is a generated preview. It is not a size recommendation and it is not a measurement of the garment.
How does the try-on sit in the shopping flow?
The button sits on eligible clothing and accessory listings inside ChatGPT, not on every product card. A shopper who never leaves the chat can preview, save, and come back. That is a different path from a retailer’s on-site virtual fitting room, which the brand controls and can connect to size charts, inventory, and returns policy.
Reference photos are saved so the user does not re-upload for the next item. Coverage of the release notes says those photos can be deleted in settings. Marketing and privacy partners should assume a shopper photo now exists in a third-party account, not only in a session on the brand’s domain. If your creative guidelines forbid using customer likenesses in ads, say so in the briefing: a try-on image the user generated is not a brand asset you can repost.
Favorites is the quieter change. A saved product is an intent signal that used to live in a retailer’s wishlist or in a browser tab. If ChatGPT holds that list, email and retargeting programs that assumed the wishlist event fired on your site will under-count consideration. Ask analytics whether assisted ChatGPT sessions are even visible in last-click reports before you judge the feature a failure.
Why does this matter for marketing teams now?
ChatGPT has been adding commerce steps through 2026, including advertising tests reported earlier in the season. Try On is a product feature, not an ad unit. It still changes the page where a shopper decides. Apparel return rates are driven by fit and appearance. A preview that looks convincing can cut returns or raise them, depending on how often the generated image mismatches the real garment. Brands will not know which until they read return reasons that mention a chat preview.
The launch is global on OpenAI’s account, but eligibility is listing-level. A brand that is absent from ChatGPT shopping results gets neither the button nor the Favorites save. Feed coverage is the first question, ahead of creative experiments.
What should marketers do this week?
Pick ten hero apparel or accessory SKUs and search them in ChatGPT on web and mobile. Record whether a shopping card appears, whether Try On is offered, and whether price and availability match the site. Gaps are a feed problem, not a branding problem.
Brief customer service and retail staff. A generated image is not a promise of color, drape, or size. Returns copy should not claim the preview is a fit tool unless your own size data backs that claim. If marketplace listings are the source ChatGPT shows, align title, primary image, and variant naming with what you want previewed.
Tell performance teams to annotate the week of October 1 in dashboards. A dip in on-site wishlist events can be a measurement move, not a demand move. Do not reallocate budget off a three-day read.
Legal and brand should confirm you are not scraping user try-on images for ads. OpenAI’s storage of reference photos is the user’s setting to manage. Your obligation is not to treat those images as licensed creative.
What should marketing watch next?
Watch whether Try On expands beyond clothing and accessories, and whether Favorites becomes a shared or exportable list merchants can see. Watch return codes and review text for mentions of a chat preview. And watch how ChatGPT shopping ads, where they are offered, sit next to an organic Try On button. A paid placement and a generated preview on the same card will need a disclosure check before the brand claims the image.
The operational date is this week’s feed test, not a future keynote. If the button is missing, the feature is not live for you yet.
When did ChatGPT Try On launch?
OpenAI announced the global launch on October 1, 2026. TechCrunch and subsequent coverage described Try On and Favorites as available on mobile and the web.
Which products can be tried on?
Eligible clothing and accessory listings. The control is a Try On button in shopping results. Users can also upload an image of an item and ask for a preview.
What model generates the image?
OpenAI said the features use ChatGPT Images 2.5, introduced on September 8, 2026.
Are shopper photos kept?
Reporting on the release notes says reference photos are saved for later try-ons and can be deleted in settings. They are not a brand asset.
Does Favorites replace a site wishlist?
No. It is a save inside ChatGPT Library. On-site wishlist events may not fire if the shopper never reaches your domain.
Is this an advertising product?
The October 1 announcement described shopping features, not a new ad unit. Brands should still check how any paid ChatGPT placements appear next to Try On.
Son Güncelleme / Last Updated: October 3, 2026
Related: OpenAI Dots agents at DevDay · Platform gatekeeping for marketers · Marketing hub
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.