What the feature does
OpenAI added a virtual try-on capability to ChatGPT, letting shoppers click a 'try on' button that now appears on clothing and accessory listings surfaced in chat. The shopper takes or uploads a photo of themselves, and ChatGPT generates images showing that person wearing the specific product. From there, users can save items to Favorites or organize them into folders inside their ChatGPT Library, the same persistent storage OpenAI has been building out across its shopping features over the past year.
The feature is live on both ChatGPT's website and its mobile app, with no stated regional limits or phased rollout beyond the initial launch. OpenAI did not name specific retail or brand partners in the announcement, and it issued no executive quote alongside the release, a notably quiet rollout for a feature this visible in the shopping experience. That restraint reads less like a minor update and more like OpenAI letting the product itself carry the announcement rather than inviting scrutiny of which merchants are or are not represented yet. OpenAI increasingly ships commerce features this way, as a line item in release notes rather than a standalone press push with retail partners lined up to comment, a pace closer to a core product team shipping weekly improvements than a retail partnerships team announcing a quarterly milestone.
The checkout retreat this follows
This launch lands shortly after OpenAI stepped back from letting ChatGPT complete a purchase directly inside the chat window, choosing instead to hand the actual transaction back to retailers' own checkout flows. That retreat suggested OpenAI had concluded the liability, fraud exposure, and merchant relationship complexity of owning checkout outweighed the convenience of finishing the purchase in-app. Virtual try-on is the opposite kind of bet: it adds a new piece of the shopping journey to ChatGPT's surface rather than removing one.
Read together, the two moves describe a company narrowing its ambitions to the parts of commerce it can execute well without taking on a retailer's operational risk. Checkout requires payment processing, fraud liability, returns handling, and tight merchant integration, infrastructure OpenAI evidently decided not to own. Try-on requires only image generation and a product catalog, a problem much closer to OpenAI's core capability. It is a retreat from operational complexity paired with an advance into visualization, not a retreat from shopping itself.
Why pre-purchase visualization matters more than it looks
A 2024 Perfect Corp. survey found that brands using virtual try-on technology saw sales conversion rates more than double, and 38.4% of respondents in the same survey said virtual fitting rooms were the ideal way to recreate physical shopping online. Those numbers explain why OpenAI is willing to build and maintain image-generation infrastructure for this specific use case even while avoiding checkout: try-on is one of the highest-leverage moments in the online apparel funnel, directly addressing the uncertainty that drives returns and abandoned carts.
For a retailer, owning that visualization moment inside your own app or website has historically been a point of competitive differentiation, something that justified in-house investment in AR and computer vision teams. If ChatGPT becomes a default place shoppers go to try on clothes before buying, retailers risk losing ownership of that moment entirely, even while keeping the transaction itself, because the uncertainty-reducing step that drives the purchase decision will have happened somewhere the retailer doesn't control and cannot instrument with its own analytics. That is a quieter form of disintermediation than losing the sale outright, but it still erodes the retailer's view into why a customer actually converted.
The competitive clock OpenAI is racing
Google has operated a virtual try-on tool since 2023, launching with women's tops before expanding to dresses in 2024 and shoes in 2025. That multi-year, category-by-category expansion gives OpenAI a clear benchmark and a clear gap to close: Google has three years of production data on what works, what shoppers trust, and which product categories render convincingly, while OpenAI is launching across clothing and accessories more broadly from a standing start.
OpenAI also already runs a shopping research feature in ChatGPT, introduced in 2025, that recommends and compares products before a purchase decision. Virtual try-on slots directly after that research step in the funnel, meaning OpenAI is now assembling a nearly complete pre-purchase experience, discovery, comparison, and now visualization, inside one interface. The only piece deliberately left out is the transaction itself, which OpenAI has chosen, for now, to hand back to merchants rather than own end to end. Google, by contrast, built its try-on tool as a standalone shopping feature rather than wiring it into a conversational assistant that already handles product research, so OpenAI's version arrives with a tighter funnel around it even if the rendering technology itself is newer and less proven at scale.
What retail CTOs should actually do about it
The near-term action is not panic about disintermediation; it is instrumentation. Retailers should start tracking whether traffic arriving from ChatGPT converts differently than traffic that used an on-site try-on tool, and whether products with existing AI-generated imagery render more reliably inside ChatGPT's try-on flow than products without it. That data will tell you within a quarter or two whether this channel deserves dedicated catalog preparation work, such as higher-resolution product photography optimized for AI image generation.
Longer term, this is a reason to treat product imagery and catalog metadata as infrastructure that serves multiple AI shopping surfaces, not just your own website. OpenAI, Google, and now TikTok are all building try-on, discovery, or agentic shopping layers that consume a retailer's catalog data in different ways. The retailers that invest now in clean, AI-consumable product data across sizing, fit, and material details will be positioned to show up well everywhere these tools surface their products, regardless of which platform ultimately wins the most usage.
