TLDR;
OpenAI has rolled out virtual try-on and product saving in ChatGPT on mobile and web. Shoppers can use a selfie to preview clothing or accessories, then save products into Favorites or folders. The generated image does not confirm fit or size, but it gives fashion brands another place where product data and imagery can shape a shortlist before the customer reaches a website.
What happened
OpenAI's release notes dated 1 October added two shopping features to ChatGPT. Eligible clothing and accessory listings can now show a Try on button. A shopper takes or uploads a selfie, and ChatGPT Images generates a picture of how the item might look on them. The same update lets people bookmark products into Favorites or organise them in folders within ChatGPT Library.
The reference photo is saved for future try-ons unless the user changes or deletes it in Settings. OpenAI's shopping documentation is clear about the limit: the generated image may not represent the product or the shopper exactly, and it does not guarantee fit or size. Measurements, product details and the merchant's returns policy still carry the decision that the image cannot.
OpenAI says the features are available in ChatGPT on mobile and web. TechCrunch described the release as a global launch and reported that the try-on experience uses ChatGPT Images 2.5. The practical change is larger than an image effect. A user can move from inspiration, to a visual preview, to a saved shortlist without opening several retailer tabs.
This sits inside ChatGPT's organic shopping experience. OpenAI says product results are selected independently and are separate from ads. Relevance can use the conversation, memory or custom instructions, while product information can come from first-party feeds, third-party metadata and public retail sources.
Why it matters
Virtual try-on changes the role of the product image. It is no longer only a packshot on a product page. It can become an input to a generated scene that helps a customer decide whether the silhouette, colour or style belongs on a shortlist. Brands with incomplete imagery or poorly separated variants may be represented inconsistently before the shopper sees the controlled retail experience.
The accuracy caveat is commercially important. A convincing generated image can raise confidence even when it cannot prove garment fit, fabric behaviour or exact colour. If the preview creates the wrong expectation, the cost appears later through returns, customer-service contacts and damaged trust. Brands should treat the try-on as assisted discovery, not a substitute for measurements, fit notes or authentic photography.
Favorites also extend the decision window. A product that is saved today can be revisited after its price, stock or variant availability changes. That makes catalogue freshness and stable identifiers more important. The brand may not control when the shopper returns, so the feed and landing page need to resolve an old shortlist into a current, purchasable option without silently switching the item.
How your brand can benefit / be affected
Start with the catalogue. OpenAI's product-feed guidance says structured feeds supply pricing, availability and seller context for discovery. Audit every apparel and accessory variant for a stable ID, current price, stock status, colour, size, material, image and destination URL. Do not collapse several colours into one record if the imagery or availability differs.
Use the product page to close the gaps left by the generated preview. OpenAI's product specification supports variant, shipping and returns information, so expose those facts consistently in the feed and on the page. Give customers a usable size guide, model measurements, fabric details, care instructions and a clear returns policy. Keep primary images colour-accurate and add views that show texture, scale and closures.
Then test the full journey with representative prompts and products. Check whether the correct variant receives the Try on option, whether saved products reopen to the same item and whether discontinued or unavailable variants fail gracefully. Compare downstream conversion and return rates for sessions arriving from ChatGPT, but do not assume the generated image caused the sale. The immediate job is to make the brand's product evidence reliable wherever ChatGPT turns it into a shopping decision.
News date: 1 October 2026. Editorial review: 3 October 2026.