Zurich's Digital Archives Are Full of Duplicate Images — and Officials, Experts and Archivists Are Finally Talking About It

Zurich's public institutions are sitting on a problem that most people never see: tens of thousands of duplicate images clogging municipal databases, research archives and cultural heritage repositories, costing money and creating real confusion about which version of a record is authoritative. The issue has moved from a technical footnote to a policy conversation, with archivists, IT administrators and data governance specialists across the city weighing in on how best to clean house.

The timing matters because several major digitisation drives are converging at once. The Stadtarchiv Zürich has been expanding its public-facing digital collections, ETH Zurich's library is deep into a multi-year project to make scientific imaging data openly accessible, and the Zentralbibliothek Zürich on Zähringerplatz recently completed a large-scale scan of its manuscript holdings. Each project has independently generated the same headache: images stored in multiple formats, under multiple filenames, sometimes across two or three separate servers, with no single flag to tell a researcher or an administrator which copy is the master.

What the Experts Are Saying

Specialists in digital preservation describe the duplication problem as structural rather than accidental. When large institutions migrate from one content management system to another — as several Zurich city departments did between 2022 and 2024 — image files routinely get copied wholesale into the new environment without a deduplication pass. The result is repositories where the same photograph or scan can exist in three or four slightly different resolutions, occupying storage space that, at current Swiss cloud infrastructure rates, is not cheap.

ETH Zurich's IT Services division, which manages storage for both the university's research data infrastructure and the ETH-Bibliothek on Rämistrasse, has publicly acknowledged the challenge of image-level deduplication in its annual data management guidelines. The guidelines, updated in January 2025, note that research datasets containing visual material are among the fastest-growing categories of stored data and that automated hash-based deduplication tools are now recommended practice for any new archive ingestion workflow.

The Stadtarchiv, housed in Neumarkt, has taken a more cautious position. Archivists there have argued that automated deletion of apparent duplicates carries real risk: two image files that appear identical at the byte level may carry different provenance metadata, different rights statements or different conservation histories. Deleting the wrong copy can erase institutional memory even when the pixels look the same. That argument has found support among digital humanities researchers at the Universität Zürich on Rämistrasse, several of whom have pointed to documented cases in European archives where aggressive deduplication led to the loss of contextually significant variant copies.

What Happens Next

The practical pressure is building nonetheless. Swiss federal archiving standards, updated under the guidance of the Schweizerisches Bundesarchiv in Bern, now require that institutions receiving public funding implement image integrity verification as part of any new digitisation contract signed after January 1, 2026. That deadline has already passed, meaning Zurich institutions that signed contracts in the first half of this year are obliged to document their deduplication methodology — or explain why they have not adopted one.

For smaller cultural institutions like neighborhood libraries in Wiedikon or Oerlikon that have digitised local photography collections with canton support funding, the burden is more practical than regulatory. Storage costs on Swiss-hosted infrastructure run significantly higher than in comparable EU markets, a gap that deduplication advocates say makes the technical fix economically straightforward to justify once the upfront audit work is done.

The conversation is also drawing in private-sector voices. Several Zurich-based firms specialising in enterprise content management, including companies operating out of the Technopark Zürich on Technoparkstrasse in the Escher-Wyss district, have been pitching AI-assisted image matching tools to public clients since early 2025, arguing that machine-learning approaches can identify near-duplicate images that byte-level hashing misses.

No city-wide mandate is currently on the table, and the Zurich Stadtrat has not formally addressed the issue as a standalone agenda item. But with digitisation budgets under review as part of the 2027 city budget planning cycle — consultations for which begin in autumn 2026 — the window for agreeing on a common standard across municipal departments is narrowing fast.