Doppl adiciona um feed de descoberta com compras integradas, construído em torno de vídeos de produtos gerados por AI (mantém-se o acrónimo em inglês)

What Doppl actually shipped today

Google Labs added a discovery feed to Doppl, its experimental app for exploring style and seeing how outfits might look on you. The feed recommends outfits, lets you virtually try items on, and — per Google — makes nearly everything you see shoppable with direct links to merchants.

Two details set it apart from a generic recommendation carousel. The feed contains AI-generated videos of real products, and the recommendations are drawn from a personalized style profile built from preferences you share and items you interact with. It is live now on iOS and Android in the U.S. for people 18 and older.

A frontend that has to reconcile generated media with real inventory

The phrase doing the heavy lifting here is 'AI-generated videos of real products.' That is a deliberate pairing: the media is synthetic, but it has to map to an item a merchant can actually sell you, because each entry carries a direct link to that merchant.

For anyone building a shopping surface, that binding is the hard part. A generated video is only useful commercially if it stays tied to a real SKU, price, and availability at the moment a user taps through. The feed's value depends on that link holding, not just on the quality of the generated clip.

Google describes the app as experimental, which is a fair label for this specific challenge — keeping synthetic product media honest against live inventory is an ongoing correctness problem, not a solved one.

The profile is the input that makes the feed personal

Doppl's recommendations are 'informed by the style preferences you share with Doppl and the items you interact with.' That is two signals: explicit preferences the user states, and implicit behavior from interaction with the feed.

Combining a stated profile with try-on interactions gives the recommender something a plain browsing history does not: a sense of what the user is willing to imagine wearing. The try-on step becomes a signal source, not just an output — every virtual fitting feeds back into what the feed surfaces next.

The implication: try-on becomes a channel, not a novelty

Virtual try-on has largely been a standalone gimmick — upload a photo, see one item. Doppl's move folds try-on into a continuous loop of discovery, fitting, and purchase, with merchant links closing the loop inside the same app.

That reframes what the try-on model is for. Here it is a component in a shopping funnel that starts with a generated recommendation feed and ends at a merchant checkout, all in the U.S. for adults as a first release. Whether the merchant links convert is the question this experiment is really testing — the try-on quality is table stakes.