AI agents · OpenClaw · self-hosting · automation

Quick Answer

ChatGPT Virtual Try-On: How It Works (October 2026)

Published:

The short answer

OpenAI rolled out virtual try-on and Favorites in ChatGPT globally on mobile and web on October 2, 2026. On supported clothing and accessory listings you get a Try On button; ChatGPT asks for a selfie or an uploaded photo and generates a preview using ChatGPT Images 2.5. It shows you how an item reads on you — it explicitly cannot confirm fit. Your reference photo is saved for reuse and is deletable under Settings → Personalization and Reference Photos. Facts verified October 3, 2026.

How to use it, step by step

  1. Find a product in ChatGPT. Ask for something shoppable — “wool overcoat under $300”, “gold hoop earrings” — and ChatGPT returns shopping results.
  2. Look for the Try On button on supported clothing and accessory listings. Not every listing has it; support depends on the merchant’s imagery.
  3. Provide a photo. ChatGPT prompts you to take a selfie or upload an existing image. A full-body photo works better than a head-and-shoulders selfie for garments; a selfie is fine for eyewear and jewellery.
  4. Review the generated preview. ChatGPT Images 2.5 renders the item on you. Expect sharper detail and faster generation than earlier image models, and expect the occasional uncanny hand or seam.
  5. Save it with Favorites if you like it. Favourited products land in your ChatGPT Library and can be filed into folders, which keeps the product and its try-on images together.

The route most people miss

You do not need ChatGPT’s shopping results at all. Upload an image of any item — including a screenshot of a product page from any website — along with a photo of yourself, and ask ChatGPT to generate the try-on. This works for merchants that are not in ChatGPT’s shopping index, which in practice is most of the internet. It is the more useful version of the feature.

What it is good at, and what it is not

ReliableNot reliable
Colour against your skin tone✅
Overall silhouette and proportion✅
Styling — does this read as “me”✅
Eyewear and jewellery placement✅
Size / fit❌ OpenAI says it cannot confirm fit
Fabric drape and weight❌ generated, not simulated
Fine detail: logos, stitching, hardware❌ frequently invented
Patterned fabric accuracy❌ patterns drift under generation

The mental model that keeps you out of trouble: this is a generated illustration of an idea, not a photograph of a fact. It answers “would I wear this” well. It cannot answer “will this fit”, “is this fabric heavy enough for November”, or “is that the actual logo placement” — because the model is drawing, not measuring.

Real virtual try-on systems in retail use AR overlays plus 3D garment models and body-shape estimation to approximate drape and fit. ChatGPT’s version skips all of that and asks an image model for a plausible picture. That is why it works on any product from any screenshot, and also why you should not trust it on size.

The privacy question, answered plainly

To avoid re-uploading on every try-on, ChatGPT persists your reference photo. That is a body or face image stored against your account.

  • Manage or delete it: Settings → Personalization and Reference Photos.
  • If you are uncomfortable with that, the feature still works — just delete the reference photo after each session and re-upload next time.
  • Standard caution applies to photos of anyone other than you. Do not upload images of other people, and particularly not children, to generate clothing previews.

Why OpenAI shipped this

Virtual try-on is the highest-intent moment in online clothing shopping and the single biggest driver of returns when it is absent. Putting it inside ChatGPT — alongside Favorites to hold a consideration set — turns an assistant into a shopping surface with a saved wishlist and purchase intent attached. For retailers and anyone doing commerce SEO, the practical takeaway is that product imagery quality now determines whether your listing is try-on-eligible, and that ChatGPT’s Library is becoming a place where consideration sets are stored outside your site.

Related: AI shopping agent vs site search, how to control AI shopping agents on your website.

Last verified: October 3, 2026.

Sources