Artprice Visual Search Opens in December and Could Help Collectors Identify a Print From a Photo

Every collector knows the moment. A print hangs in a café or turns up in a box at a flea market. No title, no signature you can read, no label on the back. You take a photo and then you’re stuck, because a text search needs words you don’t have.

Artprice, the French art market database owned by Artmarket.com, says it’s about to fix that. Artprice Visual Search opens as a global beta for all Artprice subscribers in early December 2026. The commercial launch follows in January 2027 inside its Premium and Professional plans, which the company says will cost noticeably more.

How it works

You upload an image of a work. The system returns 49 related works, each with a relevance score as a percentage. Artprice says the core model was trained on 23 million annotated reproductions of artworks gathered over almost 30 years by its historians and editors. A wider collection of about 145 million images is still being processed.

The useful part isn’t the picture matching on its own. Each image is tied to about 180 data fields: school, technique, materials, signatures, dimensions, provenance, catalogue raisonné references and auction history going back to 1700. Artprice describes a retrieval-augmented setup that compares the visual features of your photo against that metadata. Its database covers more than 930,000 artists and over 30 million auction results from 7,200 auction houses.

What it means for posters

Posters are a hard case for any visual search. The same image exists in many printings, sizes and states, and a later reprint can look almost identical to a first edition in a phone photo. A system that returns the closest 49 works may well find the image. It won’t tell you which printing you’re holding. Paper, size, the printer’s line at the bottom margin and the condition still decide that, as we went through in Collecting Vintage Travel Posters, What the Market Knows.

Where it should help is the first step, the one in Finding the Poster You Half Remember in a Collection of Thousands: going from an image to a name. Once you have the artist and the title, the rest of the research opens up. Auction results, dealer listings and museum records all need a name to start from.

Treat the company’s claims about accuracy with care until independent users have tried it. Artprice says its error rate is marginal and that it beats general search engines, but those are its own numbers. The founder, Thierry Ehrmann, has argued strongly for AI built on authored data, as Publishing House covered in Artprice Founder Thierry Ehrmann Argues for Author Control Over AI Writing. Visual Search is that argument turned into a product. The December beta will show whether 30 years of annotation beats a general model trained on everything.