ChatGPT Try-On: Upload a Selfie, Wear the Clothes Before You Buy

October 3, 2026 · 5 min read

Online clothes shopping has one eternal gamble: will this actually look good on me? On October 1, OpenAI launched two new ChatGPT shopping features worldwide — virtual try-on and a Favorites folder — that turn that gamble into a preview you can actually see.

How it actually works

In ChatGPT's shopping results, some products now carry a "Try On" button. Tap it, upload a selfie or a full-body photo, and the model generates an image of you wearing the item. Don't have a photo handy? You can even hand it a screenshot of the product itself and ask it to dress you in it. The whole loop stays inside the chat: no jumping to another app, no waiting for a brand's own AR filter to load. The model reads the garment from the product image, reads your posture from the selfie, and stitches the two together — pose, lighting, and all.

Everything runs on the newly released ChatGPT Images 2.5, which OpenAI says renders light and shadow more naturally, follows editing instructions better, and generates faster. The second feature, Favorites, is the quieter one: save products and try-on results into a Library inside the app, and over time you build a virtual wardrobe you can browse before you buy.

The returns problem

Here's the part the fashion industry should read twice: clothing has some of the highest return rates in e-commerce, because a size chart can't tell you what looks good. People order three sizes and send two back. If a selfie plus a "Try On" button can kill even a fraction of those "it looked better on the model" moments, retailers save on shipping both ways — and the planet saves on cardboard. And there's a second layer that matters to platforms: try-on data is preference data. A model that knows what you tried on knows what you almost bought.

Google did it first — sort of

ChatGPT isn't first to this idea. Google built similar virtual try-on into Photos last year. But placement matters: Google's version lived inside a photo app, while ChatGPT's sits inside the shopping conversation itself — exactly where the buying decision happens. TechCrunch reports the launch, and the subtext is hard to miss: AI shopping is moving from search to trial.

One caution before you upload

The feature asks for the most personal input there is: photos of your body. Try-on quality goes up the more the model can see — which is exactly why you should think before you upload. Stick to a neutral photo, and watch what lingers in the background or the file's metadata. Delete test photos you no longer need: a selfie uploaded for one jacket shouldn't live forever in a chat history you forget about. A quick privacy check costs nothing; regret is expensive.

What it means for you

The honest take: try-on won't kill returns — fit, fabric, and quality still can't be photographed onto you. But the direction is right. For two years, shopping AI answered "what should I buy?" Now it's starting to answer the better question: "how does it look on me?" There's a second-order effect worth watching, too: once AI knows how clothes look on you, recommendations stop being about categories and start being about fit. Retailers who plug into that shift early will sell more, ship back less, and learn your taste faster. The rest of us just get fewer surprises at the mailbox.

Stay current: we publish a daily AI news roundup — 20 stories, plain-language summaries, no hype.