OpenAI launched a global virtual try-on feature for clothing and accessories within ChatGPT on October 1, 2026. The tool allows users to upload photos of themselves to visualize how specific garments, shoes, and accessories look on their frame before making a purchase from third-party retailers.
This release marks a significant shift in OpenAI’s commerce strategy. The company has officially discontinued its “Instant Checkout” experiment, which previously attempted to facilitate direct in-app purchases. Instead, the new system focuses on discovery and visualization.
How the AI Try-On Works
The feature is powered by the ChatGPT Images 2.5 model, which OpenAI released in September 2026. This specific iteration of the model focuses on improved lighting and texture rendering, aiming to reduce the “uncanny valley” effect often seen in AI-generated overlays. By analyzing the drape of fabric and the way light interacts with different materials, the model attempts to provide a more realistic representation of how a garment fits a user’s unique body shape.
Users can interact with the shopping features through the following workflow:
- Reference Photos: Users upload a clear image of themselves to serve as a baseline for all try-on requests.
- Discovery: Users can search for specific items or upload images of clothes they find on other websites to see them rendered on their own photo.
- Organization: ChatGPT now includes a Favorites library where users can save products and organize them into specific folders for future comparison.
OpenAI has clarified that these tools are intended for style visualization only. The company warns that the virtual try-on images do not guarantee specific fit or sizing accuracy, as the AI does not yet account for the technical measurements of individual garment patterns.
Privacy Controls and Data Storage
The introduction of personal “reference photos” into the ChatGPT ecosystem has raised new questions regarding data privacy and security. By default, any photo uploaded for a virtual try-on is saved to the user’s profile to streamline future sessions. This allows the AI to remember a user’s height and build without requiring a new upload for every item.
Users who prefer not to have their images stored long-term can manage these assets by navigating to Settings > Personalization > Reference photos. From this menu, saved images can be viewed, replaced, or deleted entirely from OpenAI’s servers. It is also important to note that these photos may be subject to the user’s overall model training preferences unless opted out in the privacy settings.
Security remains a high priority for users following a report on September 28, 2026, which indicated that OpenAI agents had accidentally leaked 53 private images from ChatGPT users onto public hosting sites. While OpenAI has stated it is tightening protocols for its autonomous agents, users are encouraged to be mindful of the background and sensitive information contained within any photos they upload for virtual try-on purposes.
