By Stuart Kerr, Technology Correspondent, LiveAIWire

AI digitising cultural heritage is now moving faster than most museum visitors realise: a single gigapixel image contains roughly fifty times more detail than a standard digital camera can capture, and Google’s robotic Art Camera can scan a painting the size of a door in about thirty minutes, a job that once took a full day of manual photography. That leap in speed is why the Google Cultural Institute digitised more than 1,000 paintings in a single year after the Art Camera launched, having managed only 200 in the five years before it. Artificial intelligence has quietly become the most important tool in the effort to preserve human cultural heritage before it is lost to decay, conflict, or climate change.

The shift matters now because so much of what museums, archives, and archaeological sites hold is fragile and finite. Fire, flooding, war, and simple age destroy irreplaceable objects every year, and once a fresco fades or a manuscript crumbles, no amount of funding brings it back. AI-assisted imaging, 3D scanning, and machine learning classification are giving institutions a way to capture what exists today at a level of detail and speed that human conservators working alone could never match.

What This Means for Anyone Who Visits a Museum

If you have ever stood behind a rope barrier straining to see the brushwork on a painting, AI digitising cultural heritage projects are quietly changing that experience. Gigapixel captures let anyone with a phone zoom into a Van Gogh or a Monet closely enough to see individual brushstrokes, something previously reserved for conservators with laboratory access. Institutions including the Getty Museum have built entire public exhibitions, such as its Dunhuang Buddhist cave art show, around high-resolution digital captures made possible by this technology.

The practical upside extends beyond viewing pleasure. Researchers can compare pigment decay and stylistic detail against an artist’s full body of work without physically handling a fragile original, reducing the risk that handling itself poses to the object. For smaller regional museums without conservation budgets, AI-assisted digitisation is increasingly the only realistic path to documenting collections before deterioration takes its toll.

AI Digitising Cultural Heritage: From Painted Canvas to Shattered Frescoes

The most striking recent work in AI digitising cultural heritage extends past photography into physical reconstruction, a theme our coverage of AI in the Art Heist touched on from the authentication side. RePAIR, a four-year project backed by €2.5 million from the EU’s Horizon 2020 programme and coordinated by Ca’ Foscari University of Venice, is using a robotic arm fitted with 3D scanners and AI recognition software to reassemble thousands of fresco fragments from Pompeii’s House of the Painters at Work, shattered first by the eruption of Vesuvius in AD 79 and further broken by wartime bombing. Specialists at the University of Lausanne have worked on the same fragments by hand since 2018, and the project is deliberately running both methods in parallel to compare results.

At the Pergamon Museum in Berlin, engineers used a conveyor-belt scanning system called CultLab3D, developed by Germany’s Fraunhofer research organisation, to capture more than 6,000 images of the 2,000-year-old Pergamon Altar in 2014, stitching them into a detailed 3D model. The altar’s exhibition hall was closed for renovation at the time, so the digital model became the only way the public could examine the ancient structure in detail until it reopened. That same logic, capture what exists now so it remains accessible regardless of what happens to the physical site, is the practical definition of AI digitising cultural heritage at institutional scale.

Rebuilding What Fire and War Destroy

The 2019 fire at Notre-Dame Cathedral in Paris offered a sobering demonstration of why this matters. Conservationists used AI-assisted 3D modelling, built from earlier laser scans and photographs of the cathedral, to recreate architectural details with precision during the restoration planning process. Without that prior digital capture, some of the cathedral’s intricate stonework would have had to be reconstructed largely from memory and photographs rather than measured data.

That lesson has pushed heritage bodies to treat digitisation as insurance rather than a nice-to-have archival exercise. A site or artefact that has been captured in gigapixel detail or full 3D scan can, in the worst case, still be studied, taught, and in some form rebuilt even if the physical object is lost. At a UNESCO-hosted dialogue on AI and museums in November 2025, the director of China’s Deji Art Museum described digitising a ten-metre historical scroll and projecting it onto a large screen so visitors could step inside the scene, while museum leaders at the same event also warned that AI-driven interpretation risks flattening the oral traditions and local nuance that a single training dataset cannot capture. The same tension between capability and oversight runs through our reporting on the Louvre security gap AI could have closed, where the cost of moving too slowly on deployment was measured in stolen jewellery rather than faded pigment.

The Cataloguing Problem AI Is Quietly Solving

Away from cameras and scanners, a less visible use of AI is transforming how collections are made findable at all, a shift not unlike the one we described in AI Behind the Curtain, where algorithms took on the unglamorous production work behind a creative field. Harvard Art Museums’ AI Explorer has generated more than 71 million machine-written annotations covering close to 400,000 collection images, using computer vision models to tag entities, themes, and descriptions that would take human cataloguers decades to record by hand. Most museum collections have vast stores of objects that have never been fully catalogued simply because no institution has the staff time to do it.

Machine learning models trained to generate baseline descriptions are closing that gap, making previously invisible parts of a collection searchable and citable for the first time. That has research value well beyond the museum walls, since scholars, students, and the public can now find and study objects that would otherwise have sat undocumented in storage indefinitely.

Where the Limits Still Show

None of this technology replaces expert judgement. AI models used to classify artistic style or flag forgeries can be biased by the data they were trained on, sometimes underrepresenting regional or historically marginalised art traditions in favour of the Western canon that dominates most training datasets. A misclassification in a museum catalogue is an inconvenience. A misclassification used to authenticate or price a work has real financial and reputational consequences, which is why every major deployment of this technology still keeps a human specialist as the final check.

The institutions getting the most value from AI digitisation are treating it as a force multiplier for their conservators and curators, not a replacement for them. CultLab3D’s own project notes make the same point about the Pergamon Altar scan: the 3D model gave the public access to a structure the museum itself could not display, but the interpretation of what that structure means was left entirely to the museum’s own archaeologists and curators, not the scanning software. The technology captures detail, flags anomalies, and manages scale. The judgement about what an object means, and what its loss would cost, still belongs to people.

What Happens Next

The direction of travel is toward faster capture, cheaper equipment, and wider access. As gigapixel imaging hardware and 3D scanning tools become less specialised and less expensive, smaller institutions and even individual researchers in under-resourced regions should be able to apply the same preservation techniques currently concentrated in major Western museums. That democratisation matters because most of the world’s cultural heritage sits outside the handful of institutions currently leading this work.

The question worth watching over the next few years is whether funding and access follow the technology. AI digitising cultural heritage can now happen faster and in more detail than at any point in history. Whether that capability reaches the archaeological sites and regional archives that need it most, rather than concentrating further in institutions that already have resources, will determine how much of the world’s cultural memory actually survives.

About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.