AI-Assisted Cataloguing: Reducing Hours to Minutes

For most auction houses, cataloguing is the single most time-consuming part of the sale preparation process. Describing hundreds or thousands of lots accurately, consistently, and in sufficient detail to satisfy both buyers and legal requirements has traditionally required experienced staff and many hours of work. A new generation of AI tools is beginning to change this calculus materially.

The Speed of AI Cataloguing

The most striking development is the sheer speed at which AI cataloguing tools can process lots. Services such as AICataloguer can process approximately one hundred lots in just thirteen minutes, using between two and six photographs per lot to identify hallmarks, signatures, logos, and other identifying features. The system generates concise descriptions of up to forty words, along with artist biographies where relevant.

This represents a transformation in throughput. A task that might previously have occupied a skilled cataloguer for an entire day can now be completed in a fraction of the time, freeing staff to focus on quality control, condition reporting, and the lots that require genuine expertise.

Mobile Cataloguing

Webtron launched what it describes as the world's first AI-powered online auction mobile cataloguing system in 2025, enabling auctioneers to catalogue lots directly from a smartphone while on site at house clearances or vendor premises. This approach allows staff to photograph items, generate descriptions, and upload catalogue entries in a single workflow, eliminating the traditional sequence of photographing, returning to the office, and cataloguing separately.

For houses that source a significant proportion of their stock through house clearances, this capability has the potential to compress the cataloguing timeline significantly.

AI-Generated Titles and Valuations

Tools such as AuctionWriter go beyond description generation to produce lot titles, formatted descriptions, and estimated values from image uploads alone. While the accuracy of AI-generated valuations should be treated with appropriate caution, the title and description generation capabilities are proving useful as a starting point for cataloguers who can then refine the output based on their own knowledge and experience.

Where AI Falls Short

AI cataloguing tools are labour-saving, not labour-replacing. They cannot assess condition with the nuance that an experienced cataloguer brings. They occasionally misidentify items or periods, particularly with unusual or specialist pieces. They cannot apply the commercial judgement needed to write descriptions that sell effectively, nor can they identify restoration, repairs, or marriages that would affect value.

The technology is notably stronger for certain categories — ceramics, furniture, and decorative arts — and less reliable for specialist items such as fine art, antiquarian books, or scientific instruments, where specific domain knowledge is essential.

The most successful implementations treat AI as a drafting tool that frees experienced staff to focus on editing, quality control, and the lots that require genuine expertise.

A Practical Approach

For houses considering AI cataloguing, the pragmatic approach is to trial the technology on a category of general goods where the descriptions are relatively straightforward and the volume is high. Measure the time saved against any reduction in description quality, and assess whether the output requires significant editing before it is fit for publication. For most houses handling general antiques and collectables, the evidence suggests that AI cataloguing delivers a meaningful improvement in efficiency without an unacceptable compromise on quality.