ChatGPT Virtual Try On Launches With Favorites: What to Know

ChatGPT virtual try on arrived on October 1, 2026, as OpenAI added a “Try on” button to eligible clothing and accessory listings and introduced Favorites for saving products. A shopper can provide a selfie, view an AI-generated preview, and keep a shortlist in the ChatGPT Library. The important distinction for shoppers and retailers is what that picture can—and cannot—answer. OpenAI says it is an illustration of how an item could look, not a reliable measure of size, fit, fabric, or the accuracy of a merchant listing.

The update turns ChatGPT’s shopping results into a more visual decision space. A user can move from describing a style to looking at products, generating an image, and saving an item without losing the conversation. OpenAI’s October 1 release notes say the features are available in ChatGPT on mobile and web. They do not provide a breakdown by country, subscription tier, retailer, or item coverage. That matters in India: the global announcement does not establish that every local apparel listing will show the button.

What launched on October 1?

The new action appears on product listings for clothes and accessories. After choosing “Try on”, a shopper takes or uploads a selfie. ChatGPT Images then generates a visual preview. OpenAI also says a person can upload a picture of a garment or accessory in a conversation and ask to see how it might look on them. TechCrunch reports that a full-body photo or a screenshot of an item can be used in this workflow. The generated output is a picture for exploration, not a photograph of an actual fitting.

OpenAI’s second addition, Favorites, lets users save a product and create folders in the ChatGPT Library. The mechanism sounds simple, but it changes where a shopping decision may happen. A person who previously bookmarked a merchant page or saved an item inside a retailer’s account can now keep a shortlist inside the assistant. That does not mean a sale occurs in ChatGPT or that a retailer can see the shortlist. OpenAI’s documentation describes saving products; it does not disclose merchant visibility into saved items or any conversion results.

How to use ChatGPT virtual try on

Start with a shopping query in ChatGPT, then open a clothing or accessory result that offers the new button. Choose “Try on” and take or upload a reference photo. OpenAI says the photo is saved for future try-ons; it can be changed or deleted under Settings, then Personalization, then Reference photos. A user can also bring an item image into a conversation and ask ChatGPT to visualize it. The availability of the button on a particular result is the decisive on-screen test; the announcement does not promise it for every brand, product, or category.

The reference-photo setting is worth checking before sharing a face or body image. It means the image is part of a reusable personalization workflow, rather than merely an ephemeral input for one preview. OpenAI documents the change and deletion controls. Shoppers should review those controls and the service’s privacy information before uploading a photo, especially if it includes other people. The announcement does not by itself specify storage duration, merchant access, or all downstream data uses, so none should be assumed.

Once the image is generated, the product can be saved to Favorites or put into a folder. Folders may be useful for comparing alternatives over several sessions: workwear in one collection, festive clothing in another, or items awaiting a price or size check. However, a saved product is only a record of interest. Its price, availability, shipping cost, and return terms may change at the merchant. Shoppers should recheck the live listing before buying.

The preview is not a fitting room

OpenAI’s shopping help page explicitly warns that a generated try-on may not accurately represent a product or the person’s appearance. It does not guarantee fit or size. That is a central limitation, not a footnote. A model can generate a plausible image while missing the fabric’s weight, a garment’s cut, a seller’s sizing convention, or how an item moves in real life. A flattering picture and a correctly fitting garment are separate claims.

A practical buying sequence is to use the preview to narrow down styles, then check the merchant’s size chart, garment measurements, material description, customer photos where available, and return policy. For Indian shoppers, delivery coverage, duties on imported items, local size conventions, and final checkout price also deserve attention. Those checks follow from ordinary online shopping risks; OpenAI has not claimed that ChatGPT validates any of them. If a product image is inaccurate or a listing is stale, a generated preview cannot correct the underlying information.

The distinction also matters for accessibility and trust. A photo-based tool may help someone see a silhouette or colour combination, but a user who cannot supply a suitable reference photo may have a different experience. OpenAI has not published try-on accuracy data broken down by body type, garment type, skin tone, lighting condition, or device. Until such evidence exists, it would be misleading to claim that the feature works equally well for everyone or reduces returns. These are questions to test, not reported outcomes.

Why this matters for retailers and brands

The commercial opportunity is a new discovery step between search and checkout. If customers describe a look, see a product on themselves, and save it in a ChatGPT folder, brands may need to make product data clearer wherever shoppers encounter it. Accurate images, concise descriptions, material information, sizes, availability, and merchant links become more useful when an assistant condenses listings into a conversational result. That is an inference about the workflow, not a claim that OpenAI has changed retailer ranking rules or granted a particular brand access to shopper data.

The launch also creates a measurement gap. A shopper may assess and shortlist an item inside ChatGPT before visiting a store. Standard website analytics could see only the eventual referral or sale, not every earlier moment of consideration. The opposite is also possible: a striking generated preview may attract attention but fail to produce a click. OpenAI has not released data on usage, retailer coverage, click-through rates, sales, or return rates for the new features. Brands should measure their own traffic and conversion rather than extrapolate from a feature demonstration.

For Indian retailers, the most immediate question is less about creating a proprietary AI fitting room and more about whether their listings are discoverable and accurate across platforms. An incomplete size table or ambiguous colour name is a problem even before an AI preview is added. A store that serves local shoppers also needs clear prices, delivery ranges, and return conditions. Lapaas Voice has previously covered Nykaa’s OpenAI conversational-shopping partnership, which illustrates a different route for retailers to put advice closer to purchase. The new ChatGPT feature is a general shopping-interface change, not an announcement about that partnership.

A separate Meta shopping-research test shows that competing platforms are also experimenting with AI-guided discovery. The products and rollouts differ, so this does not establish equivalent try-on features.

There is a broader change in who owns the shortlist. A merchant-owned wishlist keeps a buyer inside a store’s account. ChatGPT Favorites stores items in a cross-product Library. That may make comparison easier for shoppers, while giving retailers less visibility into pre-purchase activity. This is a strategic implication of the described feature, not a confirmed data-sharing arrangement. The current documentation does not say retailers can see which users have favorited their products.

What independent reporting confirms

TechCrunch’s October 1 report by Sarah Perez describes the try-on button, photo-based previews, and the Favorites Library. It also places the launch after OpenAI’s earlier instant-checkout experiment, which TechCrunch says underperformed. That history helps explain the shift toward product discovery, but it does not prove OpenAI has abandoned all checkout ambitions. The present announcement is about previewing and saving products.

Fashion trade publication WWD, in an October 1 report by Evan Clark carried by Yahoo, separately describes the try-on and saving features and asks what they mean for fashion commerce. That Yahoo page is a syndication of one WWD report, not a second independent account. Search Engine Roundtable also covered the release on October 1 and verified the button, reference-photo controls, and Library folders against OpenAI’s own documentation. Together with the first-party release note, these reports establish the event date and the core functionality. They do not establish transaction performance.

A separate timing point is easy to confuse. Several launch reports mention ChatGPT Images 2.5. OpenAI introduced Images 2.5 on September 8, before the October 1 shopping update. The company says that image model brings more natural lighting, richer textures, more reliable editing, and up to 50% lower generation latency than Images 2.0. Those are OpenAI’s performance claims about the model, not independently verified measures of garment fidelity or evidence that Images 2.5 launched on October 1. The October 1 news is the shopping workflow built on existing image capabilities.

Where ChatGPT shopping stands against other options

Virtual clothing preview is not a new invention. Retailers and platforms have been testing image-based fitting and style visualization for years, and TechCrunch notes Google’s earlier virtual try-on work. ChatGPT’s distinguishing feature is the place where this step occurs: inside a general-purpose assistant that can discuss a style, return product candidates, generate a preview, and save a shortlist. That integration could reduce the effort of switching between apps, but whether it produces more confident or better purchases remains untested in public evidence.

OpenAI’s shopping help says product results are selected independently, rather than bought as advertisements. Users should still inspect the merchant link and product details. An AI-generated product label or description can be wrong, and a familiar brand name inside a conversation is not a substitute for checking the destination store. Lapaas Voice’s earlier analysis of AI shopping agents provides context for how assistants can shape choice before checkout. The new visual step increases the importance of being precise about what the assistant knows and what the retailer guarantees.

The most defensible view today is narrow: ChatGPT has gained a useful-looking visual shopping tool and a place to save finds; the feature is documented for mobile and web; the preview is explicitly not a fit guarantee; and there is no released evidence of improved conversion, lower returns, or universal product coverage. For shoppers, it is a way to explore. For merchants, it is another surface where product information needs to stand up to scrutiny. Its business impact will become clearer only when usage and purchase behaviour can be measured.

Questions to watch after the rollout

Three practical questions remain open. First, how consistently does the “Try on” button appear across brands, sizes, and countries? OpenAI’s announcement gives the broad mobile-and-web availability but no comprehensive product map. Second, how well do generated previews preserve the actual garment and the person’s appearance across diverse inputs? The help page’s warning shows why that question matters; model-quality claims alone do not answer it. Third, do saved products send shoppers back to merchants often enough to change sales? Neither OpenAI nor the publishers cited here provide launch-period figures.

Those unknowns should keep the conversation grounded. A user can experiment with ChatGPT virtual try on now, but should treat the picture as a visualization and verify the size and listing before paying. Retailers can watch whether the assistant becomes a source of qualified traffic, while avoiding claims that a generated image predicts fit. OpenAI’s latest shopping release brings the fitting-room metaphor into a chat window; the real fitting-room test still belongs to the shopper and the garment.

Sources and method: This article checked OpenAI’s October 1 release notes, its shopping guidance, and its separate September 8 Images 2.5 announcement. The event was corroborated with original October 1 reporting by TechCrunch, WWD (via Yahoo syndication), and Search Engine Roundtable. WWD’s syndicated copy is counted as one publisher. Product-performance claims are attributed to OpenAI; no sales or fit-accuracy result is implied.