eBay
GenAI generation of a product listing from a photo
With Magical Listing (2023), eBay generates a complete product listing from a photo: 30% of US app sellers tried it and more than 95% of them used the AI-generated description.
Key points
- GenAI generation of a product listing from a single photo, with seller review.
- The Magical Listing tool, eBay's in-house vision and LLM models in the Seller Hub.
- 30% of US app sellers tried it, more than 95% kept the AI description.
- Evidence B, confirmed status.
Objective
Reduce listing friction, especially for beginner sellers facing the cold-start problem, by automatically generating a complete listing.
The deployment
The Magical Listing tool generates a listing from a single photo taken or uploaded in the eBay app. The AI analyzes the image and produces a title, description, release date, category, and subcategory, and combines with eBay's other components to suggest price and shipping. The seller reviews and edits before publishing. The direct goal: speed up the creation of detailed, consistent listings, especially for new sellers. Rolled out from late 2023 in the Seller Hub and the mobile app.
Results Proof B
Adoption and retention metrics communicated by eBay and reported by specialized press (Retail Dive, TechCrunch). Usage figures at marketplace scale, but no direct financial result.
How it works
Documented architectureThe stack in detail
- llm Modeles vision + LLM eBay (in-house) Recognize the object in the photo and generate title, description, category, and attributes; eBay does not publicly detail the models.
- infra Seller Hub et app mobile vendeur Magical Listing's entry points: photo taken or uploaded, then reviewed and edited by the seller before publishing.
- outil Taxonomie catalogue et moteur de suggestion prix Existing eBay components combined with the generation: category and subcategory mapping, price and shipping suggestions.
How it runs, concretely
For ops teams-
1Taking the photo customer
The seller photographs or uploads an image of the product in the app.
-
2Analysis and generation AI
The AI recognizes the object and generates title, description, category, and attributes.
-
3Additional suggestions AI
The platform suggests price and shipping from its other components.
-
4Review and publish customer
The seller edits if needed then publishes the listing.
The photo quality and object recognition. On a poorly framed or hard-to-identify product, the generated listing goes off and the seller has to redo everything.
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
- product images
- a category taxonomy
- an attribute reference
Org prerequisites
- a vision-language model or API
- a seller validation loop
Possible stack
- a multimodal image-to-text model
- mapping to the catalog taxonomy
- a price suggestion engine
The plan, step by step
- Step 1Frame the fields to generate (title, description, category, attributes) and the target taxonomy.Deliverable: A spec for the generated listing and sets of test photos
- Step 2Wire a multimodal model (API) into the listing creation flow.Deliverable: A photo-to-listing prototype on the test catalog
- Step 3Build the mapping to the taxonomy and the seller review screen.Deliverable: An editable pre-filled listing in pre-production
- Step 4Open a beta to a segment of sellers and measure trial and description retention.Deliverable: Adoption and retention figures on the beta
- Step 5Generalize and track listing completeness and time to list.Deliverable: The feature in production with an adoption dashboard
First step: Wire a multimodal model into the listing creation flow and let the seller edit the output.
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 EBay's new AI tool generates product listings from photos Established press archive pending
- S2 eBay rolls out a tool that generates product listings from photos Established press archive pending
- S3 Selling is now even easier with AI Interested party 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.