AR Product Try-On: Which Categories Convert and Which Do Not
A founder-voice guide to AR product try-on for ecommerce: where it lifts conversion, where it falls flat, and how to pick the categories worth the build.
If you have looked at AR product try-on and you are not sure whether it pays back the build cost, the answer depends almost entirely on the category. AR is not a universal lift. It is a tool that works very well for a specific set of buyer questions and very poorly for others. This guide covers which categories convert with AR, which do not, and how to pick where to spend the build budget first.
Which product categories convert with AR try-on?
Four categories carry the bulk of the AR conversion lift in 2026: furniture and lighting (placement in the buyer's actual room), eyewear (face try-on), watches and rings (wrist or finger try-on), and cosmetics for shade matching (foundation, lipstick, hair color). These categories share one trait: the buyer's question is geometric or compositional, and AR answers it with a reliable visual. iOS Quick Look ships USDZ; Android Scene Viewer ships GLB; both render the AR step in one tap from the product page on most modern phones.
Apparel, footwear, headphones, sport equipment, and most accessories see weaker AR results because the buyer's question is about fit, materials, or use, and AR rarely answers any of those convincingly. For these categories, a real model in three body sizes, a wear-on-model composite, or a Gaussian splat captures the answer better than an AR placement.
What makes an AR-friendly category
Three properties decide whether AR converts.
The buyer's question is spatial. Furniture buyers ask "does it fit." Eyewear buyers ask "does it suit my face." Watch buyers ask "does it match my wrist size." All three are answered by a visual placement. Apparel buyers ask "does it flatter my shape," which is not spatial in the same sense.
The geometry is stable. A sofa, a chair, a watch, and a pair of glasses have predictable shapes that render cleanly in AR. A pair of headphones around a head, a t-shirt draped on a torso, a sneaker on a foot all have variable shapes that AR struggles to render with conviction.
The lighting is forgiving. AR placements look right under most ambient light if the geometry is simple and the materials are well-defined. They look wrong under variable light if the materials are fabric, leather, or skin where the rendering has to match what the buyer sees in their room. Furniture and watches survive this; apparel and footwear often do not.
The AR-friendly categories all have all three properties. The borderline categories miss one or more.
Furniture and lighting
The strongest AR category in premium DTC. A buyer at $1,500+ for a sofa, a $400+ pendant lamp, or a $2,400 dining table needs to see the product in their room before they commit. AR placement gives them that with one tap from the PDP.
The build that works:
- USDZ for iOS, GLB for Android. Both formats from the same base model, exported with AR-compatible materials.
- Under 5 MB total per file. Larger products (sofas, tables, beds) push toward the high end; smaller items (lamps, side tables, accent pieces) sit at 1 to 3 MB.
- Default fabric or finish. AR ships with the most popular variant. Updating the AR model when the buyer changes the configurator option is a v2 problem; ship the default first.
- Clear placement instructions. "Point your phone at the floor where you want the sofa" beats "tap to place." Buyers fail AR more often when they do not know where to point the phone.
The furniture configurator guide covers the broader AR+configurator setup. AR alone moves furniture conversion measurably; AR plus a configurator moves it more.
Eyewear
The category that built mainstream consumer AR. Warby Parker proved virtual try-on for glasses, and the playbook now works across most premium eyewear brands. The buyer points the phone at their face, the AR places the frames, and the buyer rotates their head to see how the silhouette reads.
Eyewear AR works because face geometry is predictable, the rendering engine handles the lighting from real cameras automatically, and the question the buyer is asking ("does this suit my face") is exactly what the AR shows. Lift on premium eyewear from AR is consistently in the 30 to 80% range for the products that ship it well.
The build is heavier than furniture because face tracking is more demanding than floor placement, but the major frameworks (Apple ARKit, Google ARCore) handle the heavy lifting.
Watches and rings
The wrist-based AR category. iOS Quick Look and Android Scene Viewer handle wrist placement cleanly with a small GLB or USDZ. The buyer points the phone at their wrist, the watch lands at scale, they rotate to see how the case sits. Same flow for rings on a finger.
This works because the question "does it fit my wrist" is geometric and the buyer can see the answer immediately. The watch product experience guide covers the broader buyer flow for premium watches; AR is one piece of it.
For necklaces and bracelets, AR is harder because the body part moves more. Most premium jewelry stores skip AR for these and use a model-on-mannequin photo composite instead.
Cosmetics: shade matching
The newest AR category to mature. AR shade matching for foundation, lipstick, eye shadow, and hair color works because the buyer is asking "what does this shade look like on me," which is exactly what the AR overlay shows.
Categories where AR shade matching converts well:
- Foundation and concealer (skin tone match)
- Lipstick and lip color (lip rendering)
- Eye shadow palettes (eye-area rendering)
- Hair color (hair tone overlay)
Categories where it does not:
- Skincare (no visual to match)
- Fragrances (no visual at all)
- Tools and brushes (the product is not on the buyer's body)
For the categories that work, the AR module from a vendor like ModiFace or YouCam usually outperforms a custom build because the rendering pipelines are mature and the makeup-specific lighting is hard to replicate.
Where AR underperforms
Five categories where AR rarely earns its build cost:
Apparel. AR for clothing has been "almost ready" for five years. The fundamental problem is that AR cannot fake fabric drape on a specific body shape under variable lighting. A buyer sees the AR and intuits that something is off. Real models in three sizes plus a fit guide consistently outperforms AR for apparel conversion.
Footwear. Foot AR works for novelty and brand awareness; it rarely closes a sale. The buyer's real question is "will this fit," and AR cannot answer that. A clear sizing guide and an honest return policy do more.
Headphones. AR for headphones often shows up as "see how they look on you," which is not the buyer's question. Audio buyers ask "how do they sound" and "are they comfortable for long sessions." A wear-on-model composite plus measurement and review data converts better. The audio PDP guide covers the broader audio-buyer flow.
Sport equipment. Bicycles, skis, racquets, and golf clubs sell on geometry and tuning, not on visual placement. A measurement chart that lets the buyer compare to their current equipment does more than AR.
Most B2B and industrial products. AR placement works only when the buyer can imagine the product in their space. Most B2B buyers care about specs, integrations, and warranty, not visual placement.
How to pick the AR investment
Walk this checklist on the product where AR is tempting:
- Does the buyer ask a spatial or compositional question pre-purchase? If yes, AR is plausible. If no, skip.
- Is the geometry stable enough to render cleanly? If yes (furniture, eyewear, watches), AR works. If no (apparel, footwear), look for an alternative.
- Does iOS Quick Look or Android Scene Viewer handle the placement reliably? Test on a real phone. If the AR is shaky or misregisters often, the build is not ready for production.
- Can the AR file ship under 5 MB? A 12 MB AR model loses the buyer at the spinner.
- Is the brand willing to maintain the AR assets? Each variant or color often needs its own AR model. Brands that cannot keep AR in sync with the catalog ship broken AR over time.
If all five answers are yes, AR is the right next investment for that product. If any are no, look at a Gaussian splat for in-page rotation, a wear-on-model composite, or a measurement-and-spec emphasis instead.
Where to start
If your category is on the AR-friendly list (furniture, lighting, eyewear, watches, rings, cosmetics shade), start with the highest-AOV product. Ship a small GLB and USDZ, test on a real phone, and measure conversion against the same product without AR for two weeks. Most stores see the lift within the first 30 days.
If your category is on the borderline list (apparel, footwear, headphones, sport equipment), AR is rarely the highest-impact next investment. Walk the Shopify PDP conversion checklist instead and ship the items higher up the list first.
The buyer is asking a specific question. AR answers it well for some categories and poorly for others. The category decides; everything else is execution.