■ MODERATE RISK ■ Creative Arts
AI is absorbing the cataloging, research-retrieval, and label-drafting work that fills curatorial back offices — but deciding what a museum should say, acquiring the objects to say it with, and owning the scholarship remains a human's job. Fewer assistant posts, same number of visions.
“AI catalogs and recommends exhibits. Your refined taste is now an algorithm.”
Our AI replacement risk score — how we score jobs
Curators are the editorial brains of collecting institutions. The work spans collection development (researching and proposing acquisitions, cultivating donors and dealers), scholarship (publication, provenance research, authentication questions), exhibition-making (concept, object selection, loan negotiation, interpretive text), and institutional life — committee meetings, grant writing, board presentations, and the diplomacy of borrowing a masterpiece from a rival institution. Behind the gallery glamour sits enormous documentation labor: catalog records, condition notes, rights management, and the perpetual backlog every collection carries.
That documentation layer is where AI has already moved in. Machine vision tags and describes objects at scale, helping institutions chew through digitization backlogs that would take staff decades; language models draft catalog entries, label copy, and translations; semantic search makes a million-object collection actually explorable; and provenance research — tracing an object through auction records and archives — is accelerating dramatically with AI-assisted document analysis. Visitor-analytics and recommendation systems now inform exhibition planning, and some institutions have experimented with algorithmically-assembled displays. The assistant-curator work of grinding through records is compressing fast.
What resists is judgment with institutional stakes. Deciding that this museum should mount this exhibition now, staking scholarly reputation on an attribution, navigating repatriation claims and community consultation, persuading a collector to donate rather than auction — these are trust, taste, and accountability functions that institutions cannot delegate to software, and audiences would revolt if they did. Museums also trade on authority: a human expert vouching for the story. Our risk score of 49 lands on a profession whose entry-level documentation rungs are automating while its senior judgment roles persist — the familiar problem that tomorrow's chief curators were supposed to train on the work AI now does.
Automatability: our editorial assessment of current and near-term AI capability
The documentation squeeze is current — AI cataloging and search tools are deployed across major collections now, and the assistant-level posts built on that labor are quietly not being refilled. Through the 2030s expect leaner curatorial departments doing more exhibitions with fewer researchers, while senior roles carrying scholarly authority, donor relationships, and repatriation judgment remain human. The profession contracts at the base and endures at the top.
It's replacing curatorial labor rather than curators: cataloging, search, label drafting, and provenance-document analysis are automating quickly, thinning assistant and research posts. The core role — deciding what a museum acquires, exhibits, and argues, and carrying the scholarly and community accountability for it — remains human because institutional trust depends on it.
Realistic but harder at the entry point. The documentation jobs where curators traditionally apprenticed are exactly what AI absorbs, so the field's chronic oversupply of qualified candidates now chases fewer junior posts. Differentiate with deep object expertise, digital-tools fluency, and public-facing scholarship rather than general museum credentials.
Mostly behind the scenes: machine-vision tagging to clear digitization backlogs, semantic search across collections, AI-drafted catalog records and translations, accelerated provenance research, and visitor analytics informing programming. Public-facing experiments — AI-curated displays, chatbot guides — exist but remain novelties; the durable adoption is in collections infrastructure.
Judgment-adjacent ones: connoisseurship strong enough to stake attributions on, donor and dealer relationships, repatriation and community-consultation competence, and the storytelling that turns collections into exhibitions people attend. Add practical command of AI collection tools — curators who direct the machines will define what the machines find.