Zara
AI-generated virtual try-on: avatar created from the customer's photos, dressed in real catalog products
Inditex announced at its 2025 annual results on March 11, 2026 that its AI virtual try-on ZARA Try-on, live since December 2025, lets customers create an avatar from their own photos and generate images of that avatar wearing real products, deployed across 43 markets with more than seven million cumulative sessions and a rollout under way to the group's other brands.
Key points
- Zara Try-on lets the customer create an avatar from their own photos and see it wearing real products.
- Launched in December 2025, the feature is deployed across 43 markets.
- More than seven million cumulative sessions, with rollout under way to the group's other brands.
- Neither conversion rate nor return rate is published: the scale is proven, the effect is not.
Objective
Remove the uncertainty about how a garment will look before purchase, which is the main friction in online fashion and the leading cause of returns. By letting the customer see themselves wearing the real product, Zara acts on conversion upstream and, potentially, on the return rate downstream. The group has published no metric on either effect: what is published is usage scale.
The deployment
Inditex put virtual try-on at the center of its 2025 annual results presentation on March 11, 2026. The feature, ZARA Try-on, has been live since December 2025: the customer creates an avatar from their own photos, then generates images of that avatar wearing real catalog products. What separates it from older image overlays is precisely that point, the product shown is the one on sale, not an approximation. The group gives two scale figures: 43 markets covered and more than seven million cumulative sessions, with rollout under way to the group's other brands. That is what takes the case out of pilot territory: a tested feature does not extend to 43 countries and to sister brands. The commercial context is published in the same document: EUR 39.9 billion in 2025 sales, of which EUR 10.7 billion online, up 4.8%. What the group does not publish, and what is missing to judge the feature: the effect on conversion and on the return rate, which are the two reasons a virtual try-on exists.
Results Proof B
The feature and its scale figures (43 markets, more than seven million sessions, extension to other brands) were announced by Inditex when publishing its 2025 annual results on March 11, 2026, and reported consistently by two press outlets the same day. Management commitment, not isolated product communication. Level B rather than A: no conversion, basket or return-rate metric is published, so the commercial result of the feature remains publicly unmeasured.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
How it runs, concretely
For ops teams-
1Avatar creation client
The customer uploads their own photos, from which an avatar is generated.
-
2Rendering generation AI
The system generates images of the avatar wearing real catalog products.
-
3Viewing and decision client
The customer views the rendering on the products of interest and decides whether to buy.
-
4Scope extension data team
The feature is extended to new markets and to the group's other brands.
The quality of catalog product visuals and the fidelity of the rendering on the avatar. An unconvincing rendering works against the objective: it increases doubt instead of removing it, and can feed returns instead of preventing them.
How your customers perceive this type of use
Sourced studiesUn ecart net separe les annonceurs des consommateurs : 77% des annonceurs voient l'IA positivement contre 38% des consommateurs (Yahoo/Publicis, 2024). Les mesures implicites confirment le rejet declare : en EEG, les pubs generees par IA produisent une activation memorielle plus faible que les pubs traditionnelles et sont decrites comme agacantes, ennuyeuses et confuses (NIQ, 2024). La disclosure a un effet ambivalent : elle augmente fortement la confiance quand elle est remarquee (Yahoo/Publicis), mais 27% des jeunes consommateurs disent faire moins confiance a une entreprise dont la pub est creee par IA (IAB, 2024).
Acceptance conditions
- Une disclosure visible : quand la mention IA est remarquee, la confiance globale envers l'entreprise augmente de 96% (Yahoo/Publicis 2024)
- Une qualite visuelle suffisante : les visuels IA de basse qualite augmentent l'effort cognitif et distraient du message (NIQ 2024)
Red lines
- Le contenu IA non declare puis identifie : 72% des consommateurs disent que l'IA rend l'authenticite difficile a etablir (Yahoo/Publicis 2024) et les marques utilisant des pubs IA sont plus souvent jugees inauthentiques ou non ethiques par les consommateurs que par les dirigeants (IAB 2024)
- Les mannequins et personnes generes par IA : 46% des consommateurs n'en veulent pas dans la publicite, l'inquietude premiere etant les standards de beaute irrealistes (Attest 2025)
Sources: Yahoo / Publicis Media (terrain Ebco) 2024 · IAB (avec Attest) 2024 · NIQ (NielsenIQ) 2024 · Attest 2025
How to replicate
Inference, not sourcedData prerequisites
- A product catalog with visuals usable for rendering, consistent from one reference to the next
- A compliant processing chain for customer photos, with limited retention
- Measurement comparing conversion and return rate between sessions with and without virtual try-on
Org prerequisites
- An explicit legal basis for processing personal photos and the avatar derived from them
- An accepted quality bar on rendering: an approximate rendering undermines the objective
- The ability to extend the feature across several brands and markets, failing which it stays a brand gimmick
Possible stack
- Image generation model conditioned on the customer's photo
- Faithful product rendering from catalog visuals
- Integration into the product page and the mobile app
- Conversion and returns instrumentation per session
The plan, step by step
- Step 1Pick the categories where rendering uncertainty blocks purchase most and where returns cost most.Deliverable: Priority product scope, quantified in return cost.
- Step 2Establish the customer photo processing chain with a legal basis and limited retention.Deliverable: Compliant avatar creation journey.
- Step 3Reach a credible rendering standard on the priority catalog before any broad rollout.Deliverable: Rendering quality validated on a sample of references.
- Step 4Open in one market with conversion and return-rate measurement, against a comparison group.Deliverable: Measured effect, not just usage volume.
- Step 5Extend market by market, then to the other brands.Deliverable: Geographic and multi-brand coverage.
First step: Instrument return measurement before opening the feature. Virtual try-on is justified by two effects, conversion and returns; if the second is not measurable by category and by size, you will never know whether the feature paid off. The Zara case illustrates that gap: the scale is published, the effect is not.
The same pattern, proven in other industries
Intuit
done-for-you genAI assistant integrated into the product to reduce friction and convert
L'Oreal (marques ModiFace)
AR virtual try-on (synthetic makeup rendering on the face)
Gatorade
self-service genAI product personalization on a D2C site
Sources
- S1 Inditex posts upbeat results for 2025 as it looks to invest in long-term future Secondary archive pending
- S2 Sales jump for Zara owner Inditex as it taps into AI for virtual changing rooms Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.