On Thursday, OpenAI globally launched virtual try-on and Favorites, two ChatGPT features meant to turn the chatbot into a more active shopping companion. Users can upload a selfie or full-body image, select clothing or accessories from ChatGPT shopping results, and tap “Try On”; products can also be saved to an in-app Library for later.¹

The rollout is powered by OpenAI’s ChatGPT Images 2.5 model, which the company says delivers “more natural lighting and richer textures,” follows editing prompts more reliably and works faster. The pitch goes beyond a digital changing room: shoppers can describe an outfit they want to build, or upload a celebrity look and ask ChatGPT to locate purchasable pieces.¹

That puts OpenAI squarely in a fashion-discovery business where Google and Pinterest already have deep roots. Google introduced a similar AI try-on tool more than a year ago, underscoring that ChatGPT is entering an established race rather than creating a new category.² OpenAI had also previously moved away from an instant-checkout concept that failed to perform well, making this a more cautious attempt to capture shopping intent.¹

But the early demonstration also exposed the central problem. In one example, ChatGPT “even dyed and styled her hair, presumably without being asked to,” the critic noted — a striking alteration when the point is to judge how an item will look on the actual shopper.²

For OpenAI, the tool is a bid to make product search visual, personal and conversational. For skeptics, that same AI creativity may undermine the promise of an honest fitting-room preview. The question is not merely whether ChatGPT can generate a stylish image, but whether shoppers can rely on it before they click buy.