The direct answer
OpenAI says ChatGPT now supports virtual try-on for clothing and accessories from product listings, alongside Favorites and Library folders that let users retain products and generated try-on images. The change matters because it links conversational discovery to a more persistent shopping consideration process. It does not prove ChatGPT has become a major transaction platform, nor has OpenAI published merchant traffic, conversion, advertising or payment outcomes for the features.
The feature is less important than the sequence it creates
Virtual try-on is familiar territory in digital retail. The more consequential detail is the combination of a product conversation, a visualisation, a saved reference photo and a persistent list of finds.
OpenAI’s 1 October release notes say a user can select Try on from an eligible clothing or accessories product listing, take or upload a selfie, and receive a virtual try-on generated by ChatGPT Images. The notes also say the reference photo is saved for future try-ons and can be changed or deleted in Settings under Personalization and Reference photos.
The same release says people can save products to Favorites or organise finds into folders in the ChatGPT Library. These features are available on web and mobile. Those are the source-bounded facts. The public release does not establish global merchant availability, a particular image model, payment functionality, conversion uplift, advertising targeting or a data-sharing model for retailers.
The important commercial question is therefore not “will this replace Amazon?” It is whether assistants begin to sit for longer between a shopper’s first expression of intent and the eventual purchase decision.
From transaction history to a consideration layer
An ecommerce platform can hold a transaction record. Search can record queries. Social platforms can record engagement signals. A conversational assistant may create another kind of environment: a place where a user describes a need, sees options, tries a visualisation, saves something and returns later.
We call that Share of AI Consideration. This is an Integrated.Social analytical framework, not an OpenAI metric and not a claim about what ChatGPT knows. It asks whether a brand has a credible chance of entering a user’s assistant-mediated shortlist while the user is still deciding.
That is materially different from claiming that an assistant knows why someone rejected every product. The release notes do not say that. They do not describe a merchant-facing preference graph, a data export, a shopper-reason model or a purchase optimisation system. The more responsible statement is narrower: persistent saved finds and a reusable reference photo could make it possible for an assistant to retain more context about a preference a user has explicitly chosen to save or reuse.
| Stage | Traditional ecommerce journey | Conversational shopping journey |
|---|---|---|
| Starting point | Ad, category navigation or keyword search | A declared need in natural language |
| Evaluation | Filters, product page, reviews and comparison tabs | Recommendations plus a conversational follow-up |
| Visual confidence | Product photography and customer imagination | Optional virtual visualisation for eligible products |
| Retention | Browser tabs, wish lists or basket | Favorites and Library organisation |
| Return moment | The shopper searches or navigates again | The user may return to a saved consideration context |
Illustrative operating model, not a product map. It describes a plausible decision journey, not an assertion about OpenAI’s undisclosed systems or commercial roadmap.
Why product truth becomes more valuable
If assistants become a larger part of evaluation, brands need public product information that remains understandable away from a single landing page. That does not mean publishing generic AI copy. It means ensuring priority descriptions, visual assets, attributes, sizes, constraints, availability and supporting evidence match what a real buyer should understand.
A virtual try-on can be compelling, but it cannot correct an unclear product proposition. A Favourite can preserve an option, but it cannot turn a vague or stale description into a trustworthy purchase decision. The commercial fundamentals remain the same: a buyer needs a clear fit, honest limits, current information and enough evidence to compare alternatives.
That is why conversational commerce belongs next to work on AI search visibility [blocked] and the broader shift toward agentic commerce and AI Mode [blocked]. The goal is not to chase every interface. It is to make the high-value public evidence behind an offer useful whenever a buyer or assistant encounters it.
A privacy boundary worth keeping clear
The presence of personal images makes precision especially important. OpenAI says users can manage the reference photo used for future try-ons by changing or deleting it in Settings. That is a user control, not a reason for a brand to assume it can access, use or infer personal visual data.
Commerce teams should resist the temptation to turn a product announcement into a surveillance claim. A retailer’s job is still to communicate its own product data, creative rights, customer permissions and offer terms responsibly. The assistant’s product controls and user data practices are separate questions, governed by the provider’s policies and the user’s decisions.
For leaders, the practical work is to clarify which data is public product evidence, which is first-party customer data and which is assistant-side personal context. Mixing them together in a strategy deck is how weak assumptions become operational risk.
Modi’s POV
Virtual try-on is not the feature I would watch most closely. Favorites is.
A saved product is a clue that a decision has not ended. The shopper may be comparing fit, price, delivery, trust, timing or an entirely different priority. Traditional ecommerce analytics often sees the click, product view, cart and purchase. It sees less of the ongoing consideration work that occurs between visits.
An assistant that can preserve a product and visual context may become part of that middle layer. That is why the new competitive question is not merely share of search. It is whether a brand appears as a credible option in the customer’s persistent consideration set.
But we should not overstate the evidence. We do not know the feature’s adoption, merchant reach, commercial terms, recommendation logic or revenue effect. “Share of AI Consideration” is a useful organising question because it tells teams what to improve: honest information, useful visuals, distinct product fit and decisions a buyer can defend. It is not a substitute for sales data.
What to do next
- Audit the public descriptions and visual assets attached to your priority products or services.
- Make constraints, eligibility, sizes, availability and proof clear enough for a buyer to compare without guessing.
- Separate product truth from personal customer data and assistant-held settings.
- Monitor the evolving commerce interfaces, but do not claim channel performance before you have a measurable test.
- Start with the free AI Answer Readiness Score [blocked] or review the scope of a paid diagnostic on Pricing [blocked].
What has not been announced
OpenAI’s public materials reviewed here do not establish that ChatGPT processes payments in this flow, provides merchant conversion reporting, supplies advertising audiences, exposes individual consideration data to brands, or guarantees recommendation placement. They also do not prove that the feature will alter the market share of any marketplace or retailer. Those are future questions, not present facts.
The brand question: are you a credible option when the conversation resumes?
For a brand, the practical challenge is not to manipulate a personal assistant’s memory. It is to make the offer coherent enough that a buyer can return to it. The visual promise, product facts, price or availability constraints, delivery terms and reasons to trust the offer should not contradict one another across a product page, a campaign and a customer-service response.
This is especially relevant in categories where the decision takes time. A person may save an item because they are comparing fit, waiting for a salary date, checking compatibility, discussing a purchase or simply deciding whether the product suits them. An assistant-mediated interface could shorten parts of that process, but no public release proves that it does so consistently. Brands should improve the decision evidence they own without assuming a new product feature will create demand on its own.
A saved item, a virtual try-on or a conversation is not a revenue outcome. It can be a useful signal to explore in a consent-aware test design, but the business should only call it commercial impact once its own measurement demonstrates an appropriate connection.
Sources
- OpenAI Help Centre: ChatGPT release notes
- TechCrunch: ChatGPT can now virtually try on clothes
- Google Search Central: product structured data
About the Author
Modi Elnadi is the founder of Integrated.Social, a London AI growth consultancy. He works with B2B and B2C teams on evidence-led AI search, product discovery and growth systems that preserve a clear line between useful experimentation and unverified commercial claims.
