ChatGPT’s Virtual Try-On: Product Data for UK Fashion Retailers

On 1 October 2026 OpenAI added two shopping features to ChatGPT: a virtual “Try on” button for clothes and accessories, and Favorites for saving products [1]. For a UK fashion or lifestyle retailer the headline is easy to read: on OpenAI’s description, a shopper can see a garment on themselves without leaving the conversation. Whether that includes UK shoppers is the open question.

The detail is less tidy. OpenAI’s pages don’t say whether the features reach UK shoppers or UK retailers’ products, and they don’t say what a product needs for try-on to work well. We build product recommendations and conversational commerce for retail and D2C brands, so we read those pages closely. This is what they say, what they don’t, and the product data we’re checking in the meantime.

What OpenAI launched

OpenAI’s release note says a “Try on” button will appear on product listings for clothes and accessories. A shopper selects it, then takes or uploads a selfie, and ChatGPT Images generates a virtual try-on. The reference photo is saved for future try-ons, and the shopper can change or delete it in Settings under Personalization and Reference photos [1]. A shopper can also upload a picture of an item in a conversation and ask for a try-on [2].

Favorites lets a shopper save any product, or create a folder to organise finds, in their ChatGPT Library [1]. Both features are listed as “Available in ChatGPT on mobile and web”.

OpenAI is careful about what try-on is. Its Help Center says try-on images “may not represent the product or your appearance exactly and do not guarantee fit or size”, and tells shoppers to check the merchant’s measurements, product details and return policy before buying [2]. We read that as a visualisation tool, not a fitting room, and it shapes the data we look at below.

The two shopping features OpenAI added on 1 October 2026. Sources: OpenAI release notes [1], OpenAI Help Center [2].

Is it available in the UK?

OpenAI’s release note and Help Center page both say Try on and Favorites are available “on mobile and web”. Neither names a country or region. Other entries in the same release notes do name the UK when a feature isn’t available there at launch, but these two entries name no country at all. Nothing on OpenAI’s pages confirms UK availability, and we haven’t confirmed it from a UK account.

The merchant side is more specific. OpenAI’s merchant page says “Shopping is currently live for ChatGPT users in the U.S.” and that it plans to expand “to more merchants and regions over time” [3]. Its FAQ adds that shopping “is currently live in the U.S. and will expand to additional regions over time” [3]. The product feed specification says the standard upload “currently targets the US” [4]. Whether a UK retailer’s product carries a Try on button in a UK shopper’s results isn’t something OpenAI’s pages settle.

On 10 November 2026 OpenAI also introduced Business Hub, where owners can claim a business in ChatGPT, connect a product feed and see mentions and website clicks. The note names the US only for the business agent, and doesn’t say which countries can claim a business [1].

How product data reaches ChatGPT

Try on appears on product listings, so the listing’s data comes first. OpenAI says product results draw on “structured metadata from first-party and third-party providers”, such as price and product description, and other third-party content [2]. In practice we see three routes in OpenAI’s documents:

• What ChatGPT can read on a retailer’s own pages. OpenAI says a product feed isn’t required, but that feeds give “greater control over how your products appear” [3].
• Shopify Catalog. For merchants on Shopify, OpenAI says product data is already integrated and “no additional work is required from individual merchants” [2]. OpenAI’s pages don’t say whether that covers UK stores.
• A direct product feed. This is built on the Agentic Commerce Protocol, or ACP. Merchants apply for access, then upload a full catalogue snapshot over SFTP, and OpenAI recommends doing so at least once a day [5].

It helps to be clear about what ACP does today. OpenAI describes it as supporting “product discovery and more accurate, personalized recommendations through shared product data”. Shoppers can discover and evaluate products in ChatGPT, but purchases are completed on merchant-owned websites or apps, and OpenAI says it is moving away from a standalone Instant Checkout [3]. So for a UK fashion retailer this is a discovery channel, not a checkout one.

The routes OpenAI describes for product data to reach ChatGPT. Sources: OpenAI Help Center [2], OpenAI merchant page [3], OpenAI Developers [4][5].

What a clothing product row can carry

OpenAI’s feed specification asks for one row per purchasable item or variant, with nine required fields: item_id, title, description, url, brand, seller_name, image_url, availability and price [4]. For fashion, the optional fields are where it gets interesting:

• Variants: each size and colour is its own row, with a shared group_id, listing_has_variations set to true, and a variant_dict such as colour and size. OpenAI says variant_dict is ignored without the other two.
• Attributes: color, size, material, gender, age_group and product_category.
• Identity: gtin and mpn, where assigned.
• Returns: accepts_returns, return_deadline_in_days and return_policy.
• Sizing: a size_system field accepts UK among other systems, but OpenAI says it needs setup and implies no size conversion. OpenAI’s spec calls these “sizing labels, not destination-country codes” [4].
• Price: written as an amount plus a currency code, for example “79.99 USD”. Which currencies a UK feed can use is “the currency agreed for your feed”, so that is a point to settle with OpenAI during onboarding.

Fields in OpenAI’s product feed specification that matter for clothing and accessories. Source: OpenAI Developers [4].

Images: what’s documented and what we check

OpenAI’s spec describes image_url as the “Main product image, showing this variant”, a direct URL such as a JPEG or PNG, with additional_image_urls for further views. It says color should be “consistent with the product image”, and that product and image URLs must be publicly accessible [4].

What it doesn’t give is a try-on rule. Neither the release note, the Help Center page nor the feed spec says what an image needs for Try on to appear or work well, or what input the feature takes from a listing. So what follows is our reading, not an OpenAI requirement.

We’d expect the listing image to matter, since it is what the shopper sees on the card. So we look at the same things we’d want for any AI surface:

• One accurate main image per colour variant, so the image, the colour name and the variant row agree.
• Clear additional views of the garment or accessory, front and back, and detail where the product has it.
• Consistent image treatment across a range, so a shopper comparing items sees like with like.
• Images and data that match the product page, since each feed row points to the product detail page for that variant.

For a quick look at a single product page, our Free GEO audit shows how AI systems read it, with results in a few minutes.

Favorites, fit and keeping data fresh

Favorites is built around coming back to a product later. OpenAI’s spec asks for item_id, group_id and offer_id to stay stable “when price, stock, title, or images change”, and for url to stay stable too [4]. For a feature where a saved product is revisited, we see why: a retired or redirected product page is a worse place for a saved item to land than a live one. OpenAI also says that when merchants change pricing or shipping terms “there may be some delay before it is reflected” [2].

Because try-on doesn’t guarantee fit, the fields that answer a shopper’s next question matter: measurements and size labels in UK terms, and a clear returns position. That’s the part of fashion retail that a generated image can’t cover.

On the site itself, the same shopper still needs help finding and choosing. We’ve covered how UK retailers do that in virtual shopping assistants, and the structured-data groundwork behind being recommended in our 90-day plan for ChatGPT.

How we approach it on the sites we build

None of this asks for a new storefront. We treat product data as the common foundation for every AI surface, so we start there. This is how we’re approaching it.

1. Auditing what’s visible today. We start with the product pages as ChatGPT can read them: price, availability, brand, images and variants in the page itself, not only in a script.
2. Tying each variant to its image. Each colour and size gets its own identity, and the main image matches the colour named in the data.
3. Keeping identity stable. Product IDs and URLs stay put when prices, stock or images change, and GTINs are included where they exist. That’s the same product-data foundation we use for AI product recommendations.
4. Putting returns and sizing in the data. Measurements, UK size labels and the returns position sit alongside the product, since try-on leaves fit to the shopper.
5. Choosing the route with the retailer. A Shopify store, a direct feed application and relying on the page itself are different routes, and OpenAI’s pages are US-first for the feed, so we set expectations on UK reach plainly.
6. Keeping the claims measured. OpenAI hasn’t published conversion or return-rate figures for try-on, so we don’t plan around any, and we watch how AI referrals actually show up for each site.

Our Free GEO audit shows how AI systems read product pages today, with results in a few minutes.

Firney is a UK web development and AI engineering company. We build fast websites, AI site search, product recommendations and conversational commerce for retail and D2C brands. Firney is a Prismic partner and a Google Cloud partner.

If it would help to talk through what this means for a particular catalogue, we’re happy to do that on a call with us.

References

1. OpenAI. “ChatGPT release notes” (entries for 1 October 2026, “New ways to shop in ChatGPT”, and 10 November 2026, “Business Hub in ChatGPT”). OpenAI Help Center. Accessed 11 November 2026. https://help.openai.com/en/articles/6825453-chatgpt-release-notes
2. OpenAI. “Shopping with ChatGPT Search”. OpenAI Help Center. Accessed 11 November 2026. https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search
3. OpenAI. “Power product discovery in ChatGPT”. ChatGPT for merchants. Accessed 11 November 2026. https://chatgpt.com/merchants/
4. OpenAI. “Products”. OpenAI Developers, Agentic Commerce. Accessed 11 November 2026. https://developers.openai.com/commerce/specs/file-upload/products
5. OpenAI. “Get Started”. OpenAI Developers, Agentic Commerce. Accessed 11 November 2026. https://developers.openai.com/commerce/guides/get-started