We Built An AI Skin Analysis Tool For Beauty Stores — Here's How It Works

Over the last several months we've been building something specific for beauty e-commerce: an AI skin analysis tool that lets shoppers diagnose their own skin from a single selfie and walk away with a personalised routine built from the store's actual catalogue. It's been running in production on two beauty stores since early this year, and we wanted to share how it works.

You can try it yourself on either of the live stores it's deployed on: sohaticare.com/pages/ai-beauty-agent or looliacloset.com/pages/ai-beauty-agent. Both run on the same engine — open one on your phone, take a selfie, see what your storefront experience would look like if your catalogue were the one feeding the recommendations.

The shopper experience, end to end

The flow starts on the product page. The shopper taps the Try-On card, enters their email, and takes a selfie directly inside the storefront — no app install, no redirect, no third-party page. Within a few seconds the AI returns a complete diagnosis: fourteen distinct skin concerns scored from 0 to 100 (fine lines, deep wrinkles, pores, dark spots, redness, dark circles, eye bags, droopy eyelids, sagging, radiance, moisture, oiliness, and two firmness gauges) plus an estimated skin age.

Each concern comes with its own gauge visualisation and a short, plain-English explanation written by the AI, tailored to the shopper's own profile — not the same boilerplate for everyone. Someone with a "normal" skin type reading gets different copy from someone whose skin scanned as oily, even if their concern scores are identical.

The shopper then ticks the concerns they actually want to address, picks a budget tier, and the AI builds them a routine. They see a Morning routine and an Evening routine written as prose paragraphs, plus a ranked product grid — top three for each selected concern — pulled from the store's live inventory, matched to their skin profile, and sorted by how well each product fits.

One tap adds to cart. The same routine and product list are also sent to their inbox so they can decide later, on a bigger screen, with the products bookmarked for them.

What the merchant actually gets

A shopper who runs this flow has already given the store three things by the time they hit the cart: an email address, a clear picture of what they're trying to fix, and a concrete budget. That's a more qualified lead than almost any other entry point on a beauty PDP, and the merchant gets it without spending a cent on an acquisition funnel.

On the catalogue side, every product the store sells is tagged against the same fourteen-concern taxonomy and the same skin-profile dimensions the diagnosis emits. So the matching isn't keyword-fuzzy — it's a direct semantic join between the customer's scan and the merchant's products. A serum tagged for "pores" and "oily skin" surfaces to customers whose scan flagged exactly that combination, ahead of more general products that only happen to mention the word "pores" in their description.

Why beauty merchants we work with are picking this up

The flat product photo in a catalogue card isn't enough information for most beauty shoppers to commit. They know they have dry patches or breakouts or fine lines, but they don't know which of the store's hundred dermo-cosmetic SKUs is actually the right one for them, specifically. Generic best-seller carousels don't solve this — they just push everyone toward the same three products.

What changes when there's an AI diagnostic in front of the catalogue is that every shopper sees a different storefront. The single mother of three with hormonal acne and a tight budget sees one set of products. The thirty-five-year-old worried about under-eye shadows sees another. The catalogue is the same; the entry point is personalised. That's the lift.

Where we are now

The skin analysis tool is live, in production, on two beauty stores — try the experience for yourself at sohaticare.com/pages/ai-beauty-agent or looliacloset.com/pages/ai-beauty-agent.

If you run a beauty or cosmetics store and you'd like to talk about how this could work on your catalogue, drop us a line.