AI search is becoming an operating surface for brand teams. The practical question is not whether a single model response changed—it is whether the change reveals something your team can verify and improve.
What changed
OpenAI’s October 1 ChatGPT release notes document two new shopping features: a Try on button on clothing and accessory product listings, and the ability to save products to Favorites or folders in the ChatGPT Library. A shopper can take or upload a selfie to generate a virtual try-on image with ChatGPT Images. OpenAI’s shopping help page also says a person can upload an image of clothing or accessories in a conversation and ask ChatGPT to show how it could look on them. The company says the features are available in ChatGPT on web and mobile.
This is a new decision step, not a new ranking rule
The update moves a supported shopping result beyond a product image, title, and outbound link: a user can create a personalized visual preview, then retain the product in a saved list. That can make a product record relevant across more than one moment in a buyer journey. It does not mean that every apparel or accessory listing has the button, that a merchant has gained a preferred position, or that a saved product will be recommended again. OpenAI does not publish an eligibility list, ranking formula, merchant participation rule, or usage data for this release.
OpenAI draws a line between product results and ads
The current shopping documentation says product results are selected independently by ChatGPT, are not ads, and are not influenced by OpenAI partnerships. It also says advertising is separate from product results. That product-design statement is important for measurement, but it is not a field test of any individual result. A brand can appear in a shopping answer, be saved to a user’s Library, show a virtual try-on control, be cited in a general answer, or run a labelled ad. Those are different events with different evidence and should not be combined into one organic visibility number.
The try-on image is not a product specification
OpenAI explicitly cautions that a try-on image may not represent either the product or the user’s appearance exactly, and does not guarantee fit or size. The practical implication is ordinary retail accuracy: shoppers still need measurements, materials, product details, images, availability, pricing, delivery terms, and returns information from the merchant. Teams should not present generated previews as proof that a particular fit claim is accurate. A visual preview can help a person explore; it does not replace the product facts that support a purchase decision or the post-purchase experience.
Reference-photo handling is part of the experience
According to OpenAI’s help page, reference photos used for try-ons are saved for future use and can be changed or deleted in Settings under Personalization. That is a user-facing control, not a merchant-data feature. Still, it makes the shopping interaction more persistent than a one-off text query: a person can reuse a reference photo and revisit saved finds later. Brands should avoid inferring personal data access, purchase intent, or conversion from that capability. The public documentation does not say what merchants can observe about a user’s try-on or Favorite activity.
What remains unknown for merchants and publishers
OpenAI’s documentation does not explain how often a Try on button appears, which catalog attributes or product sources affect eligibility, how visual previews affect clicks or sales, or whether Favorites change later result presentation. It also does not link the update to a new merchant feed requirement or a new paid-placement product. OpenAI already offers product-feed documentation for merchants seeking more current product information in ChatGPT, but the October update does not establish that a feed guarantees inclusion, a try-on control, a recommendation, or a sale. TechCrunch’s independent launch coverage confirms the product rollout and notes the broader competitive context, but it does not supply performance evidence.
A measured response for commerce teams
Start by checking the public product information a buyer would need after a preview: a stable canonical product URL, accurate title and variant, current price and availability, dimensions or sizing guidance, materials, fit notes, product imagery, shipping, returns, and support contact. Keep those facts consistent across the site and any approved catalog feed. Then observe a small set of representative shopping requests in defined account, market, and device conditions. Record the complete result, named products, displayed links, available controls, date, and any later referral or conversion data. This is a baseline for observation—not a prescription for influencing ChatGPT’s selection.
Separate the evidence record
For a release like this, a useful measurement sheet keeps four layers apart: the product result shown; the answer language or brand framing; the user-facing interaction such as Try on or Favorite; and downstream visits or commercial outcomes. Add market, device, account condition, exact request, source URL, and time to each capture. If a result changes, compare the product facts and conditions before assigning a cause. This prevents a rich new interface from being mistaken for a documented change to ranking, citations, or recommendation logic.
Bottom line
ChatGPT now lets shoppers virtually try on supported clothing and accessories and save products for later. That is a concrete expansion of the product-discovery experience, not evidence of a new way for every merchant to rank, be cited, or convert. The immediate work for brand, SEO, and commerce teams is evidence hygiene: make buyer-critical product facts verifiable, observe the experience under controlled conditions, and measure product presence, interaction, referral, and outcome as separate signals.