The Future of AI in E-Commerce: Virtual Try-On Technology

Virtual try-on spent a decade as a demo at trade shows. In the last two years it crossed into being a production tool that real shoppers use to make real decisions. The technology isn't done — but it's far enough along that the strategic question is no longer "should we adopt try-on?" It's "where in our flow does it pay back fastest, and where is it still too early?"

Where try-on actually works in 2025

Apparel with predictable drape — t-shirts, hoodies, knitwear, casual dresses — works well. The model interaction is straightforward, the rendered output is photo-realistic, and customer comprehension is high. These are also categories with high return rates, so the conversion gain compounds with the return reduction. Brands deploying try-on on these categories see the cleanest ROI signal.

Eyewear and accessories work well too, often via AR rather than full body composition. The mechanical simplicity of "thing rests on face" makes the AI's job easier, and shoppers trust the result more. The user's own face is in the loop, which is why eyewear try-on has reached mainstream adoption faster than apparel.

Cosmetics — lipstick, eye shadow, foundation matching — also belong in the working column. The shopper sees the product on themselves through their phone camera, in real time. The novelty wore off in 2023; the conversion lift didn't.

Where it still falls short

Tailored garments — suits, structured outerwear, anything where fit at the shoulder or waist matters — are hard. The AI can render the look but can't yet predict how the actual garment will sit on the actual body. Returns in these categories are still dominated by fit, and try-on doesn't help the way it does for casual apparel. If your bestseller is a structured blazer, try-on isn't your conversion lever yet.

Footwear is borderline. Visual try-on is fine for "how does this colour look with this outfit." It doesn't address the comfort and sizing questions that drive most footwear returns. The technology to predict fit from foot scans exists but isn't yet a mainstream consumer feature.

Don't oversell. If a category isn't a try-on win yet, leaving it out is more honest than half-implementing. A try-on button that produces a believable-but-misleading preview is worse than no try-on at all — it sets a return expectation the product can't keep.

What's coming next

Two threads worth watching. First, fit prediction layered on top of visual try-on — using customer body measurements to predict whether an item will actually fit, not just how it will look. This is the harder problem, and the brands cracking it will own the high-ticket apparel category. Expect 2026 to be the year fit prediction starts shipping in production.

Second, native try-on in messaging surfaces — try a piece in iMessage or WhatsApp before clicking through to the PDP. The browse-to-try friction drops to nothing. This will reshape where the try-on actually happens in the funnel — the conversion event might land at "shared a try-on with a friend" instead of "added to cart." Brands that prepare for this funnel shift will pull ahead.

The third thread, less certain: AR glasses. If consumer-grade AR glasses see meaningful adoption, in-store try-on becomes a glasses experience instead of a mirror. That's three to five years out, but worth tracking.

How to think about adoption

Don't treat try-on as a binary "we have it or we don't." Treat it as a per-category decision. For each category in your catalogue, ask: does the technology produce believable output for this product type? Does the conversion math work? Does the return rate justify the investment? If three yeses, ship try-on for that category. If two or fewer, wait.

The brands losing today are the ones that adopted try-on across their entire catalogue regardless of category fit, then quietly turned it off when the metrics didn't move. Half-shipped features train customers to ignore them.

The strategic question

Try-on isn't the disruption. The disruption is that physical retail's last real moat — "you can see it on yourself before you buy" — is now an e-commerce feature too. Brands whose store strategy assumed that moat will survive need to re-plan. Brands that lean into try-on early get to redefine what online apparel shopping feels like before competitors catch up.

Avriro's try-on tools cover the categories where it works today. Try them on your catalogue if you want to see where the gain shows up first.