When the Machine Speaks for the Collection: Deployed Generative AI in Museums and Art Galleries—A Scoping Review

Abstract

Generative artificial intelligence has begun to answer for the collection in the institution’s own voice, yet commentary has accumulated far faster than working systems have been documented, and no prior review identified by this search has consolidated and critically read the record of what has actually been deployed. Using a single open bibliographic index (OpenAlex)—chosen for reach and reproducibility, at the acknowledged cost of studies indexed only elsewhere—and a scoping protocol, this review narrows the literature to the systems genuinely placed in a working museum or gallery and observed in use, and maps them onto a function-based taxonomy. From a de-duplicated pool of more than eight thousand records, twenty-eight works describing twenty-seven distinct deployments are included. Three centers of gravity emerge: conversational guides that address the visitor, co-creative installations that make the act of generation the exhibit, and behind-the-scenes tools that read and catalog the collection. Against that map, the review weighs the evidence, confining every judgment of how well a system worked to the six studies that meet the review’s predefined evaluation-strength criteria. The result is a field whose map is clear while its proof is thin: the deployments can be named and sorted with confidence, but evidence that any delivers what it promises rests on a handful of cases, and the one production-scale system is also the only one tested at scale. Education, access, and audience understanding are almost absent from the record this search retrieved. The review closes with evidence-calibrated guidance for institutions weighing a deployment—a decision aid, a set of questions to put to a vendor, and stage-by-stage recommendations for selection, implementation, and evaluation, each marked with the strength of the evidence behind it—and an agenda redirecting the field toward the civic work it has set aside.