Google AI try-on and Doppl: what search-native try-on means for fashion brands

KEY TAKEAWAYS Summary by the editors

1. Google announced an AI virtual try-on experiment in Search Labs at I/O on 20 May 2025, letting US shoppers upload a full-length photo of themselves to see apparel listings on their own body.
2. Google said the feature uses a custom image generation model for fashion that accounts for how materials fold, stretch and drape on different bodies, and that it works across billions of listings in its Shopping Graph.
3. After a broader US launch for clothing in mid-2025, try-on was extended in October 2025 to Australia, Canada and Japan, with shoes added as a category.
4. Doppl, a standalone Google Labs try-on app launched in the US in June 2025, gained a shoppable discovery feed in December 2025, but Google Labs closed the app on 30 April 2026, with try-on continuing on listings and image results across Google.
5. For brands, search-native try-on means the product images and data they already supply to search and shopping channels become the input for a try-on they do not control, so image quality and accurate product data matter more than before.

Google has turned virtual try-on from a niche brand feature into a function of search itself. Between May 2025 and April 2026 it moved an AI try-on from a Search Labs experiment into Google Search and Shopping in several countries, launched and then closed a standalone app called Doppl, and folded the technology back into product listings. For fashion brands the lesson is that try-on is increasingly something that happens on a platform's surface, using the brand's own product imagery, rather than only on the brand's website.

What did Google launch for virtual try-on?

Google announced the feature at its I/O developer conference on 20 May 2025. In a post on its official blog, the company said a try-on experiment was rolling out that day in Search Labs in the United States. Shoppers could tap a “try it on” icon on an apparel listing, upload a full-length photo of themselves and see how the item would look on them. Looks could be saved or shared. The first categories were shirts, trousers, skirts and dresses.

Google described the technology as a new custom image generation model for fashion that understands how materials fold, stretch and drape on different bodies. It also said the feature would work across billions of apparel listings in its Shopping Graph, the product database that powers Google Shopping. In the same announcement Google described a separate agentic checkout and price tracking feature, which is not part of try-on but shows the direction: discovery, visualisation and purchase increasingly happen inside the search interface.

How did the rollout develop after launch?

After the Labs test, Google introduced AI try-on for clothes to US shoppers; TechCrunch reported on 8 October 2025, two months after that launch, that Google was extending clothing try-on to Australia, Canada and Japan and adding shoes, using the same flow: tap “Try It On”, add a full-length photo, and see the item on a digital version of yourself within seconds. The same report noted that Amazon and Walmart had introduced similar virtual try-on features, so Google was entering a crowded field rather than creating a new one.

Read also

How Zalando uses AI: the assistant, personalisation and size advice

What was Doppl and why did it close?

Doppl was an experimental app from Google Labs, launched on 26 June 2025 in the US on iOS and Android. Users uploaded a photo or screenshot of an outfit, from a shop, a friend or social media, and Doppl showed it on an animated digital version of themselves. It could also turn static images into AI-generated videos to suggest how an outfit might move. Google was candid about limits: its launch post said the app might not always get things right and that fit, appearance and clothing details may not always be accurate.

On 8 December 2025 Google said Doppl had added a shoppable discovery feed. Recommendations were based on a style profile built from the preferences users shared and the items they interacted with; nearly everything in the feed was shoppable with direct links to merchants, and the feed included AI-generated videos of real products. It was limited to US users aged 18 and over.

Google Labs then announced that the Doppl app would shut down on 30 April 2026, after which it would stop working and be removed from the app stores. Its official page framed this as concluding the Doppl experiment and pointed users to virtual try-on on product listings and image results across Google, where shoppers can click a “try it on” button. Google did not publish usage figures, so it is not possible to judge from public information how widely Doppl was used.

What does search-native try-on mean for fashion brands?

The main shift is control. A brand that builds try-on into its own site decides which products are enabled, how garments are photographed and how results are presented. When try-on runs on listings in a search engine's shopping database, the brand's existing product imagery and data become the raw material, and the platform's model decides how the garment is rendered on a shopper's body.

• Image quality becomes a try-on input. Clean, well-lit product images that show the whole garment give a generative model more to work with than cropped or heavily styled shots.
• Product data accuracy matters more. Category, colour, material and size information shape how items appear in shopping results, and errors travel further when the platform builds new experiences on top of them.
• Expectations are set before the shopper reaches the brand. If a try-on image suggests a looser or longer fit than the real garment, the disappointment may arrive as a return to the brand.
• Try-on is visual, not a size guarantee. Google's own Doppl disclaimer is a useful reminder that generative try-on shows appearance; it does not confirm that a given size will fit.

Does generative try-on reduce returns?

There is no public evidence yet that Google's try-on reduces returns, and the company has not claimed it does. Generative try-on works from a photo of the shopper and an image of the product; it does not take precise body measurements or the garment's pattern dimensions, so it is better at answering “do I like how this looks on me?” than “will size M fit me?”. Retailers that report measurable return reductions from fit technology, such as Zalando with its size advice and measurement-based fitting room pilots, rely on body measurements, return reasons and product measurements rather than images alone.

That does not make search-native try-on irrelevant to returns. A shopper who has already seen a garment on a likeness of their own body may be less likely to order several colours or styles to compare at home. But until platforms or independent researchers publish results, brands should treat it as a discovery and confidence feature rather than a returns solution.

Read also

How is AI used in fashion customer service, and where does it fail?

What should fashion brands do now?

1. Audit the product images and data you supply to search and shopping channels, with try-on in mind: full-garment views, accurate colours and correct category and material attributes.
2. Monitor where your products appear with try-on enabled in the markets where it is live, and check sample results for misleading fit or length.
3. Keep size and fit guidance on your own pages strong, because platform try-on does not replace measurement-based size advice.
4. Track returns from search and shopping traffic separately where you can, so any effect of platform try-on becomes visible in your own data.

Google's year of experimentation shows how quickly a platform can put try-on in front of millions of shoppers, and how quickly it can change the format. Brands that treat their product imagery and data as infrastructure, rather than as campaign assets, are best placed to benefit whichever surface wins.

Frequently asked questions

How does Google's virtual try-on work?

On supported apparel and shoe listings, shoppers tap a try-on button and upload a full-length photo of themselves. Google's custom image generation model then shows the item on the shopper's body within seconds. Looks can be saved or shared.

Where is Google virtual try-on available?

Google opened try-on to US shoppers in mid-2025 and, according to TechCrunch, extended it in October 2025 to Australia, Canada and Japan, adding shoes. Availability can change, so brands should check Google's current documentation for each market.

Is the Doppl app still available?

No. Google Labs shut down the Doppl app on 30 April 2026, describing it as the conclusion of the experiment. Google says try-on technology remains available on product listings and image results across Google.

Does Google try-on show whether a size will fit?

Not reliably. Google's own Doppl launch post warned that fit, appearance and clothing details may not always be accurate. Generative try-on shows how a garment may look, but it does not use precise body or garment measurements the way size recommendation tools do.

GuideThe complete guide to AI in fashion e-commerce, marketing and retailRead the complete guide