The short version: No. ChatGPT’s virtual try-on, launched worldwide on 1 October, generates a picture of you wearing an item from a photo — useful for colour and proportion. OpenAI itself warns the image may not show how a product looks or fits. Fit lives in garment measurements and fabric, which a picture cannot read.
OpenAI added a “Try On” button to ChatGPT’s shopping results on 1 October, TechCrunch reported. You upload a selfie or a full-body photo, and ChatGPT generates an image of you wearing the item you are looking at. You can also upload a picture of clothing from somewhere else — a screenshot of a retailer’s product page, say — and ask ChatGPT to try that on. A new Favorites feature saves products into folders in your ChatGPT Library, next to the try-on images.
The feature runs on ChatGPT Images 2.5, the image model OpenAI launched alongside it, which the company says produces more natural lighting, sharper detail and more realistic textures. It is available globally on mobile and web, according to Retail Gazette, which noted on 20 October that OpenAI said in September that ChatGPT had reached 1.2 billion weekly users. That puts what Retail Gazette called “a virtual fitting room” in front of a very large audience — and makes it worth being precise about what that fitting room can and cannot do.
How does ChatGPT’s virtual try-on actually work?
ChatGPT’s try-on is an image generator, not a measuring tool: it combines a photo of you with a photo of a product and produces a new, plausible picture of the two together.
What it has to work with is limited. It has your photograph — one angle, one pose, whatever you were wearing, the light in the room. It has the product image — usually a garment photographed on a model or flat on a white background, often styled, pinned or steamed for the shoot. From those two, the model predicts what the combination might look like. It does not have your measurements unless you give them, and even then it has no reliable way to turn them into drape. It does not have the garment’s measurements, its fabric weight, its fibre composition or its construction, because none of that is visible in a product shot.
OpenAI is candid about the limits. Its warning, as reported by Retail Gazette, is that the generated images may not accurately show a product’s appearance or how it will fit, and that shoppers should still check sizing information and return policies. That is the right instruction, and it is the one most people will skip.
What can a virtual try-on tell you, and what can’t it?
A try-on image is good at the questions a mirror answers from a distance, and poor at the ones only your hands and a tape measure can.
It can give you a fair sense of colour against your skin and hair — whether a camel coat warms you up or washes you out, whether a cobalt knit is too much. It can show rough proportion: a cropped jacket against a longer one, a wide leg against a straight one, where a hem might fall relative to your frame. And it can help with outfit-building — seeing whether a new piece sits with things you already own.
It cannot tell you whether the shoulders of a jacket end at your shoulder bone, whether a waistband will gape or dig, whether trousers have enough room in the seat to sit down in, or whether a knit will cling or skim. It cannot tell you how a fabric moves, creases or recovers. It cannot tell you the difference between a 300-gram wool flannel and a thin polyester blend printed to look like one, because in a photograph they can look identical. And because the image is generated to look convincing, it will tend to show you a version of the garment that fits — which is the most persuasive thing it could show, and the least informative.
Why doesn’t a picture show fit?
Fit is a relationship between two sets of numbers and one material, and a photograph carries none of the three.
The first set of numbers is yours: chest, waist, hips, shoulder width, inseam, the length from the nape of your neck to where you want a coat to end. The second is the garment’s: its flat measurements, which differ from the size on the label and differ between brands that print the same size. The gap between the two is ease — the room a garment has around your body — and ease is what makes a coat comfortable over a jumper or a trouser sit cleanly without pulling.
The material decides what that ease does. A heavy wool drapes and holds a line; a light viscose falls and clings; a knit with elastane stretches and recovers; a rigid denim does not move at all until it has been worn in. Two garments with identical measurements in different fabrics fit differently. Our guide to reading a clothing composition label explains how fibre percentages predict this before you ever touch the piece.
None of those numbers is in a product photo. Some of them are on the product page, if you scroll — which is where the useful work happens.
How should you use AI try-on before buying clothes online?
Treat the try-on image as a first filter for taste, then do the checks that actually decide whether the piece will work.
One. Use it for colour and proportion only. If the image tells you a colour is wrong for you or a length looks off, that is useful and probably right. If it tells you something fits, ignore that part.
Two. Find the garment measurements, not just the size chart. A size chart tells you what body the brand designed for; a garment measurement table tells you what the piece itself measures. Lay a similar item you own and like flat, measure it across the chest, the shoulders and the length, and compare. A difference of two or three centimetres across the chest is the difference between a jacket you wear and one you return.
Three. Read the composition and weight. Fibre content and, where given, fabric weight in grams per square metre tell you how a piece will hang and how long it will last. For outerwear, our guide to judging a wool coat sets out the numbers that matter.
Four. Check the return policy before you buy, not after. OpenAI’s own advice. Know the window, who pays for return postage and whether refunds come as money or credit.
Five. Decide what happens to your photo. ChatGPT saves the reference photos you upload for future try-ons, according to Retail Gazette; you can change or delete them in ChatGPT’s personalisation settings. If you would rather not leave a full-body image stored, delete it when you are done.
Is ChatGPT the only virtual try-on tool?
No — Google launched its own virtual try-on in the UK in December 2025, letting shoppers see clothing from many product listings on their own photo, including dresses, jackets and footwear, Retail Gazette reported. Many fashion retailers also run their own try-on or fit tools on their sites.
What changes with ChatGPT is where the try-on happens. Because you can upload a screenshot from any shop, the image no longer depends on a retailer having built anything; the assistant becomes the fitting room for the whole web. That is convenient. It also means the image is made by a company that has no information about the garment beyond its picture, and no responsibility for whether it fits.
It is also worth remembering that this is not OpenAI’s first push into shopping inside ChatGPT. TechCrunch reported in March that its plans to make ChatGPT more like Amazon, built around instant checkout, were not going well; try-on is a more persuasive feature, built to make buying feel easier. Easier buying is not the same as better buying.
Where the real fitting room still is
The pieces most worth owning are, almost by definition, the ones a try-on image is worst at judging, because their value is in cloth and construction rather than in how they photograph. An independent designer or craft workshop will usually give you exact garment measurements and answer an email about how a piece fits a particular build; a good vintage or estate seller will measure an item for you; accessible-luxury brands increasingly publish fabric weights and detailed measurement tables. Mid-tier mass-market fashion, which publishes the least and changes its blocks most often, is where a try-on image is most likely to flatter something that will not last — skip the mid-tier mass market entirely and the problem largely disappears.
Frequently asked questions
Is ChatGPT’s virtual try-on free?
TechCrunch and Retail Gazette describe it as a global launch on mobile and web, but neither sets out whether it differs by ChatGPT plan. Check inside the app for what your account includes.
Can ChatGPT try on clothes from any website?
Yes, in the sense that you can upload a screenshot or image of an item from another shop and ask ChatGPT to try it on, according to TechCrunch. The result is a generated image based on that picture, not on the retailer’s product data.
Does ChatGPT keep my try-on photos?
It saves your reference photos for future try-ons, Retail Gazette reported. You can change or delete them in ChatGPT’s personalisation settings.
Is virtual try-on accurate for sizing?
Not reliably, and OpenAI says so: it warns that generated images may not accurately show how a product will fit and advises checking sizing information and return policies. Compare garment measurements with a piece you own instead.
What is the best way to check fit when buying clothes online?
Measure a similar garment you already like, compare it with the product’s flat garment measurements, read the fibre composition and fabric weight, and buy only where the return terms are clear.
The honest reading
ChatGPT’s try-on is a useful way to rule things out: the wrong colour, the wrong length, the shape that was never you. It is a poor way to rule things in, because the one thing it cannot see — how a particular cloth, cut to particular measurements, behaves on a particular body — is the thing you are paying for. Use the picture to narrow the list. Use the numbers to decide.
The next move is yours.
Sources: TechCrunch, “ChatGPT can now virtually try on clothes for you” · Retail Gazette · TechCrunch on ChatGPT’s instant checkout
