Virtual Try-On Ads: Where the Format Works in D2C

A shopper stalls on a fitted tee for the same reason they stall on anything bought sight-unseen: the one thing a product photo cannot answer is whether it will sit right on their body. Virtual try-on ads promise to close that gap inside the ad unit. The shopper sees the product rendered on themselves, or on a model built to their proportions, before the click.

The format is AR layered onto a shoppable ad - eyewear, watches, sneakers, makeup, and now apparel, pulled from a product catalog and rendered in the feed. Meta first opened AR Try On to select merchants in 2019 and now lets advertisers feed an AR product catalog into Dynamic Ads so the try-on effect attaches to whichever product a person is shown, per Meta’s own developer note.

The apparel brands running it built it off the ad unit

ASOS launched a hybrid try-on across around 10,000 products in February 2026, letting shoppers upload their own photo or pick an AI-generated model in their likeness, per ASOS. It runs inside the ASOS iOS app, not a Meta placement.

Zara built the same idea into its own app - a virtual fitting room that dresses a 3D avatar generated from a face and a full-body photo, per its 2026 launch coverage. Warby Parker popularised eyewear try-on inside its own app years earlier and is now leaning back toward physical stores, sunsetting its home try-on program in favour of showrooms and its in-app tool.

The pattern holds across all three. The brands with the most try-on volume own the surface it runs on. Meta sells the format as a buyable placement; the reference implementations live in first-party apps, where the brand controls the render, the load time, and the data.

What the format is actually good at

AR try-on works best where the product is rigid and the fit question is trivial. Meta’s own expansion piloted APIs with ModiFace and Perfect Corp - both beauty engines - because a lipstick shade or a sunglasses frame renders cleanly onto a face.

FARFETCH’s watch try-on, built with WANNA, drove +47% site visits, +22% add-to-bag, and +81% add-to-wish-list against comparable watches without it over a 14-week window, per WANNA’s case data. The same writeup flags that the test was not A/B controlled and the sample was small - trends, not proven results. Numbers this soft are common in try-on case studies, and the widely repeated “30% lift” attached to FARFETCH is a separate figure from its Advantage+ Shopping campaigns, not its try-on.

Fit is the hard case, and fit is the whole pitch

The objection virtual try-on ads sell against is fit, and fit is exactly where the rendering breaks. How a garment drapes, stretches, and gathers is a material-physics problem that remains unsolved, per 2026 technical surveys. The systems land silhouette and proportion long before they land fabric behaviour.

So the format is strongest for the products that never had a fit problem and weakest for the ones that do. A fitted tee is the hard case.

Fresh Clean Threads answers that same fit objection without an AR unit at all. Its Google account runs search ads segmented by body type - short men, tall men, big guys, DadBod - with 7 of its 25 search ads carrying a dedicated headline and, in some cases, a dedicated landing page, per our teardown.

The teardown’s read: Google captures “someone at the exact moment they search t-shirts for short men,” which Meta structurally can’t. The fit signal FCT buys through a keyword is the same signal try-on ads try to manufacture through a render.

The returns claim, and how far it stretches

The commercial case for try-on rests on fewer returns. Independent 2026 estimates put the reduction anywhere from 15-20% for flat 2D overlays to around 40% for full AR, per one technical survey. The spread is wide enough that the number tells you more about the implementation than about the format.