Zurich's Digital Archive Crisis: The Key Decisions Ahead on Duplicate Image Replacement

Zurich's major cultural and civic institutions are facing a concrete reckoning over duplicate digital images — redundant files clogging servers, inflating storage costs, and undermining the integrity of public archives. The pressure to act has reached a tipping point, with Stadt Zürich's Stadtarchiv, ETH-Bibliothek, and the Kunsthaus Zürich all independently confirming ongoing audits of their digital holdings this year. The question is no longer whether to act, but how — and who pays for it.

The issue gained momentum after a working group convened in early 2026 under the umbrella of the Schweizer Bibliotheksverbund to examine standards for deduplication across cantonal and municipal collections. Participants noted that the explosive growth in digitisation projects since 2020 — accelerated partly by pandemic-era closures that pushed institutions online — had left legacy systems stuffed with near-identical image files, variant scans, and orphaned metadata. Managing that redundancy is not a vanity project. Storage infrastructure at the scale maintained by institutions on Rämistrasse and the ETH-Zentrum campus on Rämistrasse 101 carries real operating costs measured in tens of thousands of francs annually.

What the Audits Have Found — and What They Haven't

Duplicate image replacement sounds straightforward: identify redundant files, designate a canonical version, retire the rest. In practice, the process is laden with institutional, legal, and technical complexity. A scan made in 2019 may carry different licensing metadata than a higher-resolution version produced in 2023 from the same physical object. Replacing one with the other without human review risks destroying provenance records that archivists and researchers depend on. The Stadtarchiv, which holds photographic collections dating to the nineteenth century and maintains a public reading room at Neumarkt 4, has reportedly flagged this problem as its primary obstacle to automated deduplication.

ETH-Bibliothek, one of the largest public scientific libraries in Europe with a digitised collection that spans historical maps to contemporary research imagery, has been piloting AI-assisted comparison tools since late 2024. The tools, developed partly in collaboration with ETH Zurich's Institute for Machine Learning, can identify visually similar images with a reported accuracy rate above 90 percent — but edge cases still require specialist sign-off. The library has publicly acknowledged that full remediation of its visual archive, estimated to contain several hundred thousand image records, will take more than a single budget cycle to complete.

The Decisions That Will Define the Outcome

Three choices now sit on the table for Zurich's institutions, and the answers will set a precedent that smaller cantonal archives across German-speaking Switzerland are watching closely.

The first is funding structure. A coordinated cantonal programme, potentially channelled through the Kanton Zürich's Fachstelle Kultur, would pool costs and technical resources. The alternative — each institution funding its own deduplication independently — risks duplication of effort and incompatible standards. Budget discussions for the 2027 cantonal cycle begin in autumn 2026, making the next four months critical for any institution that wants the issue on the agenda.

The second decision concerns which image wins when duplicates conflict. Institutions must agree on a hierarchy: resolution, date of scan, rights status, or metadata completeness. No Swiss-wide standard currently exists. The Schweizerisches Sozialarchiv at Stadelhoferstrasse 12, which manages one of the most-used photographic collections in the German-speaking world, has been drafting internal guidelines since March 2026 that could serve as a template if adopted more broadly.

The third is public access continuity. When a canonical image replaces its duplicates in a public-facing database, existing permalinks often break. Researchers who have cited specific archive URLs in published work find their references dead. Ensuring redirect infrastructure is built before, not after, deduplication runs is a technical step that has been skipped in comparable projects elsewhere in Europe — with damaging results for academic credibility.

A coordinating session is scheduled for September 2026 in Zurich, hosted by the Schweizer Bibliotheksverbund, where institutions are expected to present their audit findings and propose a shared framework. Observers say that session will be the real test of whether Zurich's institutions can align — or whether each will retreat to solving the problem alone, at greater cost and with less coherent results.