ChatGPT now has a virtual try-on button for clothing and accessories, but it is a shopper feature that lives only inside ChatGPT: a brand cannot embed it in its own store, measure it or control it. OpenAI added the “Try on” button to product listings in ChatGPT shopping on October 1, 2026, on web and mobile, generating the preview with ChatGPT Images from a selfie the shopper takes or uploads.

For merchants the headline is less about the feature than about where it sits. Shoppers can now see themselves in a product before they ever reach a retailer’s site, and the retailer sees none of it. This report compares five virtual try-on options that a brand can run on its own store or build into its own product, ranked on whether the brand keeps the experience, the data and the checkout path, and on published pricing we checked directly on October 17, 2026.

There is no single best option; it depends on what you sell and who builds it. WearFits is the most complete store-owned choice because one vendor covers shoes, bags and apparel. Genlook is the cheaper, apparel-focused route with a free plan and conversion analytics. TryOn API is the right layer for a team that wants to build its own experience on top of many models rather than commit to one.

What ChatGPT’s try-on does, and what it does not

The launch is well documented. SEO Roundtable, Let’s Data Science and TechAIWire all date it to October 1, 2026. A “Try on” button appears on eligible clothing and accessory listings; the shopper provides a selfie; and ChatGPT Images generates a preview. Shoppers can also upload a screenshot of a garment from any store and ask how it would look on them. A companion Favorites and Library feature lets them save products into folders.

The limits matter more to merchants than the capabilities. OpenAI says the images “may not represent the product or your appearance exactly and do not guarantee fit or size.” Genlook’s analysis of the launch found that the feature works only inside ChatGPT, does not appear on a store’s own product pages, and does not appear in merchant analytics, and it found no merchant opt-in, opt-out or control. Purchases still complete on the retailer’s own site.

The data question

ChatGPT saves a shopper’s reference photo for future try-ons, which can be replaced or deleted under Settings, Personalization, Reference photos. Reporting on the launch adds that uploads from personal accounts can be used for model training by default unless the shopper turns off “Improve the model for everyone.” We could not confirm how the setting is handled for workplace plans. For a brand with its own privacy promises, that is a reason to prefer an experience where it chooses the model, the retention policy and the consent flow.

Visual preview is not fit data

A generated image answers “how might this look on me,” not “what size should I buy.” That is true of ChatGPT and of every generative apparel tool in this report, which is why the honest framing for a merchant is complementary: try-on lifts confidence in the look, while sizing tools such as those covered in our broader try-on platform ranking address fit. Footwear and bags are different, because AR and 3D rendering at true scale can show proportion and size rather than just style, which is where WearFits’ shoes and bags coverage differs from apparel-only apps.

Where this is heading

Conversational try-on is becoming a default surface: Google already runs one inside Search, Shopping and Gemini, and ChatGPT has now followed. Both are closed surfaces, so the practical response for a brand is not to wait for access but to make its own product pages the better place to try something on, with analytics it controls. Merchants who want to build that themselves can call the same class of models through a router such as TryOn API; those who want it installed this week can use a store app.