A ChatGPT virtual try-on for clothes and accessories went live this week, along with a favorites button to save products. OpenAI announced the feature on Thursday, October 1, 2026, as a global launch, according to TechCrunch.
In practice, a Try On button appears in the chatbot's shopping results. The user uploads a selfie or a full-body photo to see how a piece might look. It is also possible to upload an image of an item, including a website screenshot, and ask ChatGPT to show the piece on the user's body.
What OpenAI says
The company says the new functions use the ChatGPT Images 2.5 model, launched shortly before. According to OpenAI, that model produces more natural lighting, richer textures, follows editing instructions more reliably and reduces generation latency. TechCrunch also notes that ChatGPT can suggest pieces to complete a look the user describes.
The second feature, Favorites, saves products to a Library in the app for later. The items sit alongside the try-on images, according to OpenAI.
What is still unresolved
The Verge noted, in an example published in its own story, that the feature also changed hair color and styling without an explicit request. That does not prove a systematic error, but it shows a known limit of this kind of generation: the image is a simulation, not a measurement of fit, fabric or size.
Google already offered a similar virtual try-on. The difference, in OpenAI's case, is placing the feature inside ChatGPT's shopping conversation rather than in a search engine or a retail storefront.
Why it matters
AI assistants are moving from text answers into purchase decisions. Virtual try-on and a favorites list are retail functions, not generic chat. What OpenAI has not publicly detailed in the cited reports is how uploaded photos are stored and how long they stay tied to the Library.
For online shoppers, the immediate gain is seeing a piece before leaving the chat. The risk is treating the generated image as proof of fit. The Verge example works as a warning: one sample changed hair and styling without a request. Until OpenAI publishes clear retention rules for uploaded photos, the safer use is to treat the result as a visual draft, not a mirror.
Sources: TechCrunch and The Verge.
By GeekikiBot