Online shopping has solved the problem of finding almost anything with a few taps. But one frustrating question has remained: what will it actually look like on me?

OpenAI is now using ChatGPT to tackle that problem.

The company has introduced a virtual try-on feature that allows shoppers to see how clothing and accessories could look on them before making a purchase. The feature is part of ChatGPT's expanding shopping experience and uses ChatGPT Images to generate a visual preview based on a user's photo.

For online shoppers, the idea is simple: instead of looking at a product photo featuring a model and trying to imagine how the item might look on your own body, you can upload a photo of yourself and ask ChatGPT to visualise the item on you.

How ChatGPT's Virtual Try-On Works

When ChatGPT displays an eligible clothing or accessory in its shopping results, users can see a Try on button.

Selecting the option allows the shopper to take or upload a selfie. ChatGPT then generates an image showing the person wearing the selected item.

But shoppers are not limited to products discovered through ChatGPT's shopping results.

Users can also upload a picture of clothing or an accessory directly into a conversation and ask ChatGPT to show what it could look like on them.

That means a screenshot from an online fashion store, a picture sent by a friend or even an image of an outfit someone spotted elsewhere can become the starting point for a virtual try-on.

The experience essentially turns a normal ChatGPT conversation into a lightweight digital fitting room.

The Technology Behind It

The feature arrives shortly after OpenAI introduced ChatGPT Images 2.5, its latest image-generation model.

OpenAI says Images 2.5 produces sharper details, more natural lighting and richer textures, while doing a better job of preserving subjects from reference photos and following editing instructions across multiple turns.

The company also says generation latency has been reduced by up to 50% compared with Images 2.0.

Those improvements matter for virtual fashion.

A useful try-on system cannot simply place a photograph of a shirt over a person's body. The generated image needs to account for the person's pose, appearance, lighting and the visual characteristics of the clothing.

Images 2.5 is designed to be better at working with reference photos, making it particularly suited to this kind of experience.

ChatGPT Is Also Becoming a Shopping Wishlist

Virtual try-on is only part of OpenAI's latest shopping push.

ChatGPT now allows users to save products to Favorites and organise saved items into folders inside their ChatGPT Library.

That means a shopper can discover a product today, save it and return to it later instead of having to search through an old conversation or start the shopping process again.

The company is effectively adding the basic functions of a shopping wishlist directly into the chatbot.

And the shopping experience can go beyond finding individual products.

A user can describe a particular style and ask ChatGPT to find products that could create that look. They can also upload an image of an outfit and ask for similar items.

That creates a different shopping journey.

Instead of starting with a product name, shoppers can start with an idea.

"I want an outfit like this."

"Find something similar but cheaper."

"Show me how this would look on me."

"Save the ones I like."

The entire process can happen within one conversation.

But Virtual Try-On Is Not a Fitting Guarantee

There is an important limitation shoppers should understand.

A generated try-on image is a visualisation, not a measurement tool.

OpenAI explicitly warns that the generated image may not accurately represent the actual product or the user's appearance. More importantly, the image does not guarantee that the clothing will fit or that the selected size will be correct.

A dress can look perfect in the generated image and still arrive too tight, too loose or differently shaped from what the AI showed.

Shoppers should therefore continue checking:

• Size charts

• Body measurements

• Product descriptions

• Material information

• Seller reviews

• Return and refund policies

The virtual try-on should reduce visual uncertainty, not eliminate the need to check actual product specifications.

Your Photo Becomes Part of the Experience

There is also a privacy consideration.

To create a virtual try-on, ChatGPT needs a reference photo of the user.

OpenAI says reference photos are saved so users can reuse them for future try-ons. Users can manage those photos through Settings → Personalization → Reference photos, including changing or deleting them.

That convenience comes with an obvious trade-off.

Users get a faster shopping experience because they do not need to upload their photograph every time, but they should also understand that their reference photo is being retained for this functionality.

For users who are particularly sensitive about personal images, that is an important setting to know about.

ChatGPT Is Moving Further Into Commerce

The bigger story may not actually be the virtual fitting room.

It is what the feature says about where ChatGPT's product strategy is heading.

Chatbots traditionally answered questions.

Shopping search engines helped people find products.

E-commerce platforms handled transactions.

OpenAI is increasingly trying to bring those stages together.

A shopper can describe what they want, discover products, compare options, visualise an item on themselves, save favourites and continue refining the search — all within ChatGPT.

OpenAI's shopping documentation also says product results are selected independently and are not advertisements or influenced by OpenAI partnerships, while some eligible products and merchants can support checkout directly within ChatGPT.

That puts ChatGPT closer to being a shopping assistant than a traditional search box.

The Fashion Industry Has More to Gain

For fashion retailers, virtual try-on could address one of the biggest problems with online shopping: uncertainty.

Customers cannot physically touch a product or stand in front of a mirror when buying online.

The more accurately technology can help shoppers visualise a product, the more information they can have before clicking "buy."

But there is another side to the equation.

If AI-generated previews become highly convincing, retailers will need to make sure their actual product photos, measurements and descriptions are equally accurate.

Otherwise, the gap between what shoppers see virtually and what arrives at their door could become another source of disappointment.

ChatGPT's virtual try-on feature is less about making AI generate prettier fashion images and more about removing friction from online shopping.

The traditional process is:

Find it → imagine it on yourself → guess your size → buy it → hope it looks right.

ChatGPT is trying to change that to:

Find it → see it on yourself → compare → save → buy.

That may seem like a small change, but it could make ChatGPT considerably more useful as a shopping destination.

The catch is that AI can show you what an outfit might look like — it still cannot guarantee what will arrive in the package.

For shoppers, the best approach is therefore to treat virtual try-on as a visual decision-making tool, not a replacement for measurements and seller information.

And for OpenAI, the move signals something bigger: ChatGPT increasingly wants to sit between consumers and the products they buy.