ChatGPT Now Has a Try-On Button. Are Your Product Photos Ready for It?

On October 1 OpenAI added a "Try on" button to ChatGPT's shopping results. It shows up on clothing and accessory listings. A shopper taps it, uploads a selfie or full-body photo, and ChatGPT generates an image of them wearing the item. TechCrunch reports it launched globally, on web and mobile, the same day as a new Favorites library for saving products.

That puts a fitting room inside the same chat where people are already asking what to buy. If you sell apparel, jewelry, bags or eyewear on Shopify, the question is what your products look like when ChatGPT puts them on a stranger.

The honest answer is that nobody outside OpenAI knows exactly how the try-on picks its source image. What we do know is enough to act on.

What OpenAI launched

From the ChatGPT release notes and launch coverage:

• The Try on button appears on clothing and accessory listings in ChatGPT shopping results.
• The shopper's reference photo is saved for future try-ons and can be changed or deleted under Settings, Personalization, Reference photos.
• Products can be saved to Favorites, or to folders in the ChatGPT Library, next to the try-on images.
• Shoppers can also upload a screenshot of any item, from any store, and ask ChatGPT to try it on.

It runs on ChatGPT Images 2.5, which OpenAI says gives "more natural lighting and richer textures" and follows editing instructions more reliably, per Dataconomy's report.

Notice the screenshot option. Your products don't even need to be in ChatGPT's shopping results to get tried on. A shopper on your product page can screenshot it and paste it into a chat. Either way, the image you published is what goes in.

Why the source image matters so much

Virtual try-on is image editing. The model takes a picture of a garment and redraws it onto a person's body. If the garment is half-hidden by a model's crossed arms, cropped at the hem or styled under a jacket, the model has to guess at the missing parts, and those guesses end up on your customer's body.

OpenAI hasn't published image guidelines for its try-on. Google has, for the apparel try-on it launched in Search, and it's the best public reference for what these systems need. Google's Merchant Center requirements ask for:

• images at least 512 x 512 pixels, "ideally 1024 pixels or higher"
• "one garment on one front-facing model or mannequin in a simple pose, or laying flat on a surface"

Google's image best practices add the details: show the entire garment, model facing forward "with arms down to the side," no hands, handbags or accessories covering the garment, no heavy wrinkles on flat lays, sleeves rolled down, hoods down.

I'd assume OpenAI's model struggles with the same things Google's does. Both are solving the same problem with similar technology. That's an inference, not a documented fact, but it costs nothing to shoot for it.

Where most Shopify catalogs fall short

Look at your main product image the way the model will. Plenty of apparel stores lead with a lifestyle shot: a model mid-stride on a beach, the shirt half-tucked, a tote bag over the shoulder. That's great for Instagram. For try-on it's a poor source, because half the garment is styled away.

Other common problems:

• The first image is a detail crop (fabric texture, a logo close-up) instead of the whole item.
• Every color variant shares the same image, so a try-on of the "olive" variant starts from a photo of the black one.
• Images are low resolution because they were compressed years ago and never re-uploaded.
• The garment is shown in an outfit, so the model can't tell where your product ends and the styling begins.

The variant problem is the one I'd fix first. OpenAI's product feed spec describes the main image field as the image "showing this variant." If your Shopify variants don't have their own images, every feed built from your catalog will repeat the same photo for every color.

Favorites stretch the timeline

The Favorites library matters for a less obvious reason. A shopper can save your jacket on Tuesday and come back the next week. When they do, the price and stock had better still be accurate.

That's ordinary product feed hygiene, but saving makes it more visible. A product that was in stock when it was saved and shows as sold out a week later loses the sale, and possibly the shopper's trust in what ChatGPT showed them.

The traffic lands on your product page

When a ChatGPT shopper clicks through, they skip your homepage. Shopify said on its Q2 call that half of all AI-referred sessions land directly on a product page, 2.5 times the rate for traditional search, per TechCrunch.

So the person arriving has already seen your product on their own body, in ChatGPT. Your product page has to confirm what they saw: the same color, the same cut, a clear size guide, and a returns policy that makes ordering feel low-risk. A try-on image is a generated guess, not a fit guarantee, and shoppers know that. Your size chart has to do the rest.

What to do

You don't need a new photoshoot for every product. Start with your top sellers and work down.

1. Make the first image a clean, full-garment shot. Front-facing on a model or mannequin, or a flat lay. Move lifestyle shots to positions two and three.
2. Give every color variant its own image. In Shopify, assign a variant image to each color so feeds and agents get the right photo.
3. Upload at 1024 pixels or larger on the shortest side. Shopify serves resized versions to shoppers, so a big original costs you nothing on page speed.
4. Keep the product unobstructed. No crossed arms, no bags over clothing, no jacket hiding the shirt you're selling.
5. Fill in the attributes. Color, size, material, gender and age group are optional fields in OpenAI's feed spec, and they're exactly what shoppers ask about. Put them in Shopify's product fields or metafields rather than only in the description.
6. Write alt text that describes the item ("women's olive linen button-down shirt, relaxed fit"), not the vibe.
7. Make your size guide text, not an image, and link it near the add-to-cart button.

Try-on images will get better whatever you do. The source photo is the one part you control. Give the model a complete, clear picture of the product and it has less to invent.