À l'intérieur de GlamAR's AI Stylist: Comment fonctionnent les recommandations hors pair

Lien copié

GlamAR's AI Stylist gives customers smart product suggestions based on what they've tried on, their skin tone, and their behaviour on your platform. It turns a single product try-on into a personalised shopping session - recommending complementary products, complete looks, and items suited to the customer's individual profile, all within the same experience.

‍

This post explains how it works, what signals it uses, how it fits into a brand's e-commerce experience, and what it delivers commercially.

The Problem It Solves

Most e-commerce experiences work like this: a customer lands on a product page, tries on or views an item, and then either buys it or leaves. The transaction is one item, one decision, one interaction.That is a missed commercial opportunity for two reasons.

‍

First, customers who are actively engaged - who have just virtually tried on a lipstick shade, a piece of jewellery, or a clothing item - are at peak purchase intent. They have already found something they like. That is exactly the moment to show them what goes with it, what else they might like, or what other customers with similar profiles bought.

‍

Second, fashion and beauty purchases are rarely isolated. A customer who tries on a blush shade wants to know what lipstick complements it. A customer who tries on a necklace is thinking about whether it works with what they already own. A customer who tries on a pair of sunglasses might be interested in a hat that completes the look. Without an AI Stylist, those connections have to be made by the customer alone - browsing, searching, filtering. With one, they are made instantly and served contextually.

‍

This is what GlamAR's AI Stylist is built to do.

What GlamAR's AI Stylist Actually Is

GlamAR's AI Stylist is a post-try-on recommendation engine built into the GlamAR platform. It delivers smart product suggestions based on three inputs:

  • What the customer just tried on
  • Their detected skin tone (from GlamAR's AI Skin Analysis, where integrated)
  • Their browsing behaviour and interaction history on the brand's platform

The result is a set of personalised product recommendations served immediately after - or alongside - the try-on experience. Customers don't have to search for complementary products. The AI Stylist surfaces them.

‍

GlamAR also has a dedicated "Looks" experience - a curated demo store where complete outfit and look combinations can be explored. For fashion brands, this extends the AI Stylist from a single-product recommendation engine into a full look-building tool, where customers can discover and virtually try on complete outfits rather than individual items.

How the Recommendation Engine Works

Step 1: The try-on creates the signal

When a customer uses GlamAR's virtual try-on - for makeup, jewellery, eyewear, clothing, or accessories - that interaction generates a signal. The system records what was tried on, in which colour or variant, for how long, and what actions the customer took (did they save it? compare with another product? return to try a different shade?).

‍

This interaction data is the starting point for the AI Stylist's recommendation logic. It knows what the customer found interesting enough to try on. That is more reliable behavioural data than a page view or even a product save - a try-on is an active, deliberate engagement.

Step 2: Skin tone calibration

Where GlamAR's AI Skin Analysis is integrated alongside the AI Stylist, the recommendation engine incorporates the customer's detected skin tone into its logic. This matters most in beauty - recommending a foundation shade that complements the customer's specific undertone, a blush that reads correctly on their skin type, or a lip colour that flatters their particular complexion rather than just returning the platform's bestsellers.

‍

Without skin tone calibration, product recommendations in beauty are generic. A customer with cool undertones and a deep skin tone is not well-served by the same foundation recommendation as a customer with warm undertones and a light complexion. The AI Stylist, informed by skin analysis data, can differentiate.

Step 3: Behavioural pattern matching

Beyond the individual try-on session, the AI Stylist learns from broader behavioural patterns - what products the customer has browsed, which try-ons they have done in previous sessions, which items they have added to cart or saved, and which they have skipped over. This builds a profile over time that makes subsequent recommendations more accurate.

‍

For a returning customer, the AI Stylist uses their history to refine what it surfaces. A customer who consistently tries warmer eyeshadow tones will not be recommended cool-toned palettes. A customer who regularly browses minimalist jewellery will not be served statement pieces.

Step 4: Catalogue mapping

The AI Stylist maps its recommendations to the brand's own product catalogue - not a generic product database. Every recommendation it surfaces is an item the brand actually sells and has in stock. This is what makes the recommendations commercially useful rather than aspirational.

‍

The brand configures their catalogue in the GlamAR Console, including product categories, complementary product relationships, and any manual overrides (for example, always pair a specific lipstick collection with the matching liner). The AI Stylist works within these parameters.

Step 5: Serving the recommendation

Recommendations are served at the point of maximum intent - immediately after or alongside the try-on experience. Placement options include:

  • Post-try-on modal: After completing a try-on, the customer sees "You might also like" or "Complete the look" suggestions before returning to the product page
  • Inline within the try-on: As the customer tries on a product, complementary suggestions appear alongside the AR view
  • On product pages: The AI Stylist can power the "Recommended for you" section on product pages, replacing static bestseller lists with personalised suggestions

What "Complete the Look" Actually Means

The most commercially valuable AI Stylist use case for fashion and beauty brands is the "Complete the Look" format - where a customer's try-on of one item triggers a curated set of complementary products that form a complete look.

‍

‍For beauty brands: A customer tries on a lipstick shade. The AI Stylist surfaces a blush that complements it, a highlighter that works with their skin tone, and a setting powder that is relevant to their detected skin type. Instead of a single product purchase, the customer sees a complete face look using the brand's own products.

‍

‍For jewellery brands: A customer tries on a necklace. The AI Stylist recommends the matching earrings, a bracelet in the same metal, and a ring that works with the same aesthetic. The try-on becomes the entry point to a coordinated jewellery set rather than a single-piece transaction.

‍

‍For fashion brands: A customer virtually tries on a dress. The AI Stylist suggests a belt that works with the silhouette, shoes in a complementary colour, and a bag that fits the aesthetic. The customer can try on each suggested item in sequence, building and previewing a complete outfit within the GlamAR experience.

‍

GlamAR's dedicated "Looks Demo Store" illustrates this full look-building approach - where complete outfit combinations are curated and each item in the look can be individually tried on. For fashion brands wanting to move beyond individual product try-on into full outfit discovery, this is the format to understand.

The Commercial Impact

Basket size

The most direct metric. A customer who discovers and tries on three complementary products in a single session is more likely to add multiple items to cart than a customer who views one product page and leaves. AI Stylist recommendations that surface relevant complementary products at the moment of peak engagement directly increase average order value.

Session depth

Customers who receive personalised recommendations during a try-on session spend more time on the platform. Each recommendation is a new try-on opportunity, which extends engagement and keeps the customer in an active shopping mindset rather than passive browsing.

Return rate reduction

When customers buy products that are recommended based on their specific skin tone, behavioural preferences, and complementary relationship to what they have already chosen - rather than just the bestsellers or editorial picks - they are more likely to be satisfied with what arrives. Accurate personalisation reduces the "this wasn't what I expected" returns that drive up operational costs for fashion and beauty brands.

Repeat purchase rate

A customer who discovers a complete look through the AI Stylist and buys multiple complementary products has a different relationship with the brand than one who buys a single item. They have engaged with the brand's curation, not just a product. Customers with higher basket depth on a first purchase typically return at higher rates.

AI Stylist vs. Standard Product Recommendations

Most e-commerce platforms have a "You might also like" or "Customers also bought" module on product pages. These work from purchase co-occurrence data - items that have historically been purchased together. They are useful. But they have three significant limitations:

‍

‍They are not try-on-aware. A standard recommendation engine doesn't know the customer just spent two minutes trying on a specific shade of lipstick. The AI Stylist does - and it uses that signal.

‍

‍They are not skin-tone-aware. A standard recommendation engine cannot calibrate a beauty suggestion to the customer's specific skin tone. The AI Stylist can, where AI Skin Analysis is integrated.

‍

‍They are generic across users. A standard recommendation engine serves the same "top recommendations" to most customers. The AI Stylist personalises to each customer's interaction history and physical profile.The AI Stylist is not a replacement for standard recommendation modules - it is a complementary, more powerful tool that operates specifically within the virtual try-on context.

How to Set Up GlamAR's AI Stylist for Your Brand

Step 1: Configure your catalogue in GlamAR Console

Log into the GlamAR Console and ensure your product catalogue is fully mapped - all SKUs, product categories, variants, and availability. The AI Stylist can only recommend products that are in the catalogue. Incomplete catalogue data produces incomplete recommendations.

Step 2: Define complementary product relationships

For each product category, configure the relationships the AI Stylist should understand. For beauty: which blush families complement which lip colour families. For jewellery: which collections are designed to be worn together. For fashion: which categories form complete looks (tops → bottoms → footwear → accessories).GlamAR's Console allows both automated relationship inference (the system learns from co-try-on data over time) and manual relationship overrides (the brand explicitly defines which products should appear together).

Step 3: Integrate with AI Skin Analysis (recommended for beauty brands)

For beauty brands, connecting GlamAR's AI Skin Analysis to the AI Stylist produces the most accurate recommendations. The skin tone and skin condition data from the scan informs the Stylist's shade and product type suggestions. Integration is configured within the GlamAR Console - no separate SDK implementation is needed if AI Skin Analysis is already deployed.

Step 4: Configure placement and display format

Decide where AI Stylist recommendations appear: post-try-on modal, inline alongside the AR view, or on product pages. For most brands, starting with the post-try-on modal is the quickest way to test impact - it appears immediately after the highest-intent moment in the customer journey without requiring changes to the product page layout.

Step 5: Track the right metrics

Set up GA4 event tracking to measure AI Stylist impact specifically:

  • Recommendation click-through rate: What percentage of customers who see AI Stylist recommendations click on at least one?
  • Average order value: AI Stylist session vs. non-AI Stylist session: The primary commercial metric - are sessions that include AI Stylist recommendations producing higher basket sizes?
  • Items per order: AI Stylist vs. standard: Are customers who engage with AI Stylist recommendations adding more items to cart?
  • Return rate: AI Stylist-assisted purchases vs. standard: Are personalised recommendations reducing returns?

Conclusion

Standard e-commerce product pages are designed for customers who already know what they want. Virtual try-on extends that - it helps customers discover what they want. The AI Stylist extends it further - it connects what the customer has discovered to everything else that could complete their look or serve their style.

‍

For fashion and beauty brands, this is not an incremental improvement. It is a different kind of shopping experience. One where a single try-on becomes a curated session. Where a customer who came to try one lipstick shade leaves with a complete face look. Where the platform works like a personal stylist rather than a product catalogue.To explore GlamAR's AI Stylist and see how outfit and look recommendations work in practice, visit the Looks Demo Store or contact the GlamAR team to discuss how it can be configured for your catalogue.

‍

Related reading:

Ready to add AI Stylist recommendations to your virtual try-on experience? Talk to the GlamAR team or explore the Looks Demo Store.

‍

Questions fréquemment posées

Article récent

30 octobre 202611 min de lectureGlamAR Magento Integration Guide for 2026Learn how GlamAR virtual try-on integrates with Magento through a code snippet or SDK, with realistic steps, timelines, and requirements now.23 octobre 202611 min de lectureGlamAR WooCommerce Integration Guide for 2026Learn how GlamAR virtual try-on integrates with WooCommerce through SDK and API, with realistic setup steps, timelines, and requirements now.16 octobre 202613 min de lectureHow Much Does Virtual Try-On Cost? 2026 Pricing GuideBudgeting for virtual try-on? See what actually drives AR try-on pricing, the common pricing models, and how to estimate a real 2026 budget.29 septembre 202616 min de lectureAR Ads sur TikTok et Instagram: un guide pour les marques en 2026TikTok AR annonces et Instagram AR annonces drive 35-40% plus d'engagement que les formats statiques. Voici comment chaque plateforme fonctionne et comment planifier une campagne.22 septembre 202611 min de lectureCréation de contenu 3D pour le commerce électronique: un guide pratique pour les marques en 2026Guide de création de contenu 3D ecommerce pour les marques : showrooms, 360 vues et try-on. Comment construire une stratégie, choisir des formats et mesurer le ROI.15 septembre 202610 min de lectureComment les événements virtuels aident les marques à lancer des produits et à conduire des ventes en 2026Les événements de lancement de produits virtuels convertissent jusqu'à 10 fois le taux de commerce électronique standard. Voici comment ils fonctionnent, quand les utiliser et comment en planifier un.8 septembre 202610 min de lectureQu'est-ce qu'un magasin de pop-up virtuel, et pourquoi les marques les utilisent-elles en 2026?Une boutique pop-up virtuelle est une vitrine 3D temporaire construite pour une campagne spécifique. Voici comment le format fonctionne, ce qu'il coûte et comment en planifier un.3 septembre 202610 min de lectureVirtual Try-On pour les sacs à main : comment AR change la catégorie des accessoires en 2026Comparer les plates-formes d'essai virtuel de sac à main par la qualité de modélisation 3D, la méthode d'intégration, le catalogue à bord, le support des appareils et le ROI commercial. B2B guide acheteur pour 2026.31 août 202612 min de lectureComment optimiser les performances du WebAR mobileOptimisation des performances WebAR pour les essais : budgets de performance, compression, chargement paresseux, démarrage de la caméra, niveaux d'appareil et mesure.21 août 202610 min de lectureComment fonctionne l'essai virtuel pour les vêtementsComment virtuel essayer sur les vêtements fonctionne, expliqué: méthodes de capture, préparation de vêtement, drap de tissu, limites d'ajustement, options de livraison, et la confidentialité.14 août 202612 min de lectureAI Analyse de la peau Vie privée et consentement : ce que les marques doivent savoirLa confidentialité de l'analyse de la peau de l'IA dépend du consentement, de la rétention et des rôles de processeur. Voir le flux de données, la liste de contrôle de conformité et les questions de l'acheteur.3 août 20266 min de lecture9 Meilleurs bijoux virtuels Essayez SDKs pour les marques de commerce électroniqueDécouvrez les meilleurs SDK qui transforment la façon dont les marques de commerce électronique offrent des essais de bijoux virtuels.