OpenAI launches virtual clothing try-on in ChatGPT with AI-generated personalized shopping previews
OpenAI has launched virtual try-on for clothing and accessories directly inside ChatGPT, allowing users to generate personalized images showing how selected products could look on them before leaving the shopping experience.
A new “Try on” option can appear on eligible clothing and accessory listings. Users take or upload a selfie, and ChatGPT Images generates a new image combining their appearance with the selected product. The reference photo can then be reused for subsequent try-ons rather than uploaded again for every item.
The October 1 update also adds Favorites and folders for products inside the ChatGPT Library, giving users a persistent place to save and organize items discovered during shopping conversations.
The result is a broader commerce workflow inside ChatGPT: product discovery, comparison, personalized visualization and product organization can now operate within the same conversational interface.
··········
CHATGPT VIRTUAL TRY-ON AT A GLANCE
........
Feature | Implementation |
Supported category | Clothing and accessories |
Main entry point | “Try on” button on eligible product listings |
Personal reference | Selfie taken or uploaded by the user |
Image generation | ChatGPT Images |
Alternative workflow | Upload an image of clothing or an accessory directly in chat |
Reusable reference photo | Yes |
Reference-photo controls | Settings → Personalization → Reference photos |
Saved products | Favorites |
Organization | Folders inside ChatGPT Library |
Platforms | Mobile and web |
Launch date | October 1, 2026 |
........
Virtual try-on is integrated with ChatGPT's existing shopping results rather than presented as a separate application. A product discovered during a conversation can therefore move directly into a personalized image-generation workflow.
··········
USERS CAN TRY ON PRODUCTS DIRECTLY FROM CHATGPT SHOPPING RESULTS
The standard workflow begins with a clothing or accessory product displayed in ChatGPT.
Selecting Try on opens the personalization process. The user supplies a selfie, and ChatGPT Images produces a generated preview intended to approximate how the item could look on that person.
OpenAI also supports a second route that does not depend on a ChatGPT product listing. Users can upload an image of clothing or an accessory directly into a conversation and ask ChatGPT to visualize it on them.
That makes the feature applicable to products encountered elsewhere. A screenshot or product image can become the starting point for a try-on request without requiring the same item to appear first in ChatGPT's shopping carousel.
The generated result remains an AI-created visualization rather than a physical fitting simulation. OpenAI explicitly notes that the image may not reproduce either the product or the user's appearance exactly and does not guarantee sizing or fit.
Measurements, sizing information, material details and merchant return conditions therefore remain separate parts of the purchasing decision.
··········
REFERENCE PHOTOS MAKE PERSONALIZED SHOPPING REUSABLE ACROSS MULTIPLE PRODUCTS
The initial selfie can be stored as a reference photo for future try-ons.
This changes the workflow from repeated one-off image generation into a reusable personalization layer. After the first setup, users can move between different clothing products without repeatedly providing another photograph.
Reference photos are managed through:
Settings → Personalization → Reference photos
Users can replace an existing reference photo or delete it.
The feature is particularly relevant when several alternatives are being compared. Instead of viewing generic product photography on different models, a shopper can generate multiple previews against the same personal reference.
Consistency of the reference image does not make separate generations perfectly comparable: generative models can still introduce changes in pose, proportions, fabric representation or other visual details. The persistent reference nevertheless removes one source of variation from the workflow and reduces the amount of user input required for repeated try-ons.
··········
FAVORITES AND FOLDERS ADD PERSISTENCE TO CHATGPT SHOPPING
OpenAI has paired virtual try-on with a second change: products can now be saved to Favorites or organized into folders inside the ChatGPT Library.
A user can bookmark a product when it appears and return to it later instead of reconstructing the original search or finding the previous conversation.
Folders make the system more useful for purchases involving multiple alternatives. Products could, for example, be separated by category, occasion or purchasing shortlist while remaining accessible from Library.
This creates a shopping sequence with several persistent stages:
Search → compare → try on → save → revisit
Previously, conversational product discovery could produce useful recommendations without necessarily creating a structured collection of the items the user wanted to retain. Favorites and folders add that missing persistence layer.
The combination with virtual try-on is more significant than either feature in isolation. A product can be discovered through conversation, visualized personally and preserved for subsequent comparison without requiring an external wishlist during the early stages of the decision.
··········
CHATGPT IS COMBINING PRODUCT SEARCH WITH GENERATIVE IMAGE PERSONALIZATION
Virtual try-on connects two systems that previously addressed different tasks: shopping search and image generation.
ChatGPT shopping can surface products with imagery, descriptions, prices and merchant options based on the user's request and available product information. ChatGPT Images can then transform a selected product into a personalized visual representation.
The resulting architecture can be summarized as:
........
Stage | ChatGPT function |
Intent | User describes the product or style required |
Discovery | ChatGPT surfaces relevant products |
Evaluation | Products can be compared by characteristics and price |
Visualization | ChatGPT Images generates a personalized try-on |
Retention | Product is saved to Favorites or a folder |
Purchase path | User can continue to a merchant or use supported checkout options where available |
........
The important technical change is that image generation is no longer isolated from the commerce context surrounding the image.
The system already knows which product the user selected and has access to the personal reference supplied for the try-on. The generated image therefore becomes another interface for evaluating a product rather than a standalone creative output.
··········
VIRTUAL TRY-ON REDUCES VISUAL UNCERTAINTY BUT DOES NOT SOLVE FIT
There are two different problems in online fashion purchasing that can easily be conflated.
The first is visualization: whether a color, silhouette or general style appears suitable when associated with a particular person.
The second is physical fit: whether a specific garment will actually fit that person's measurements, body proportions and preferences when manufactured in a particular size.
ChatGPT's new feature primarily addresses the first problem.
A generated preview can make an abstract product photograph more personal, but it does not replace accurate size charts, garment dimensions, fabric behavior or physical fitting.
This distinction also affects how the images should be interpreted. Generative reconstruction can alter details, including the precise shape of a garment, how material falls, product proportions or aspects of the person's appearance.
For consumers, virtual try-on is therefore best understood as an additional visual decision layer, positioned between generic product imagery and the purchasing decision rather than as evidence of exact physical fit.
··········
THE FEATURE EXTENDS CHATGPT FURTHER INTO THE SHOPPING DECISION PROCESS
ChatGPT has progressively expanded beyond answering product questions into a more structured shopping environment.
Users can already ask for products according to criteria such as budget, characteristics or intended use; compare alternatives; inspect product information; and follow links to merchants. Some eligible products and merchants can also support checkout functionality within ChatGPT.
Virtual try-on adds a type of information that conventional product search cannot provide directly: a generated representation conditioned on the individual shopper.
Favorites and folders then preserve the results of that discovery process for later use.
This creates a materially different interface from a conventional search engine result page. Instead of repeatedly reformulating searches, users can maintain a conversational context containing preferences, shortlisted products and personalized visualizations.
The effectiveness of that model will depend on the reliability of the product information and generated imagery, particularly in a category where relatively small differences in cut, proportions and material can influence a purchasing decision.
··········
OPENAI IS TURNING CHATGPT SHOPPING INTO A CONTINUOUS MULTIMODAL WORKFLOW
The October update moves ChatGPT shopping closer to an end-to-end decision environment.
A user can describe what they want, receive product options, compare alternatives, select an item, generate a personalized try-on with ChatGPT Images, save the product and return to it later from Library.
The strongest addition is not simply the generation of another AI image. It is the connection between product data, conversational context, a personal reference image and persistent product organization inside one workflow.
Virtual try-on still cannot determine whether a garment will physically fit or guarantee that the generated representation matches the real product exactly. Its role is narrower: reducing the gap between seeing a product on a standard listing and imagining how it could look on the individual considering it.
For ChatGPT, that adds another stage of the purchasing process to an interface increasingly designed to handle product discovery, evaluation and personalization without forcing each step into a separate application.
··········
FOLLOW US FOR MORE.
DATA STUDIOS
datastudios.org
