How Paris's Digital Archives Ended Up Full of the Same Image Twice: The Long Road to a Duplicate Problem

Paris's municipal digital infrastructure is carrying a quiet burden. Across the Bibliothèque de la Ville de Paris, the Atelier Parisien d'Urbanisme's public asset portals and the sprawling media archive assembled for the 2024 Olympic Games, duplicate images now account for a significant share of stored files — slowing systems, inflating server costs and, in some cases, presenting contradictory versions of the same historical document to the public. The problem did not arrive overnight.

Understanding why matters today because the city is at a pivot point. Grand Paris Express construction documentation, Seine riverbank regeneration photography and the post-2024 legacy activation archives are all being consolidated into a single interoperable platform that the city's Direction des Affaires Culturelles has been developing since late 2024. If duplicate content is not resolved before migration, archivists warn, the new system inherits the old disorder at scale.

A Decade of Competing Platforms

The roots go back to roughly 2013, when the Mairie de Paris accelerated digitisation across its departmental libraries. Each arrondissement mairie was encouraged to upload local heritage photography independently, with little cross-referencing between databases. The 4th arrondissement's Centre Pompidou-adjacent municipal archive and the 13th arrondissement's digital collection — built partly around the Paris Rive Gauche urban renewal project — each ingested material from overlapping photographic agencies. By 2018, internal audits flagged duplication rates inside certain thematic folders exceeding 30 percent, according to planning documents reviewed by The Daily Paris.

The 2024 Olympics accelerated the pile-up. Paris 2024's official media operation generated an enormous volume of photography from venues including the Trocadéro, the Stade de France in Saint-Denis and the temporary arenas along the Champs-Élysées corridor. That material was distributed to at least four separate municipal or semi-public bodies — including Paris Musées, the Régie Autonome des Transports Parisiens communications unit and the tourism authority Choose Paris Region — each of which archived locally without a shared deduplication protocol. The result was thousands of near-identical frames sitting across disconnected servers by January 2025.

Rue de Rivoli's municipal data offices and the tech teams based at the Hôtel de Ville have known the duplication problem was structural rather than incidental for at least three years. What changed recently is cost. Cloud storage pricing for the city's primary vendor contract, renewed in early 2025, rose by an estimated 22 percent compared with the previous cycle, according to budget documentation presented to the Paris City Council's digital affairs committee in March 2026. That made the previously abstract problem suddenly legible in euros.

What Deduplication Actually Requires

Solving it is harder than it sounds. Paris's archives contain not just exact pixel-for-pixel copies but near-duplicates — the same photograph cropped differently, watermarked for different institutional uses or converted between file formats over successive migrations. Standard deduplication software flags exact matches; near-duplicates require human review or more expensive machine-learning tools. The Bibliothèque nationale de France, based in the 13th arrondissement on Quai François Mauriac, has been running a near-duplicate detection pilot on its Gallica platform since 2023 and its technical findings have been shared informally with city archivists, though a formal joint protocol has not yet been adopted.

The consolidation project targeting a unified platform is scheduled for a phased rollout beginning in the fourth quarter of 2026. Before that deadline, archivists at the Direction des Affaires Culturelles face the unglamorous work of reviewing flagged batches, establishing which version of a duplicated image holds the highest resolution and correct metadata, and retiring redundant files without accidentally destroying the only surviving copy of something genuinely unique. That last risk is not hypothetical — at least one early batch deletion in 2022 removed what turned out to be the sole digital scan of a pre-Haussmann street map held at the Carnavalet museum on Rue de Sévigné.

For the public, the practical implication is straightforward: anyone using Paris's open-data image portals for research, journalism or civic projects should download and independently verify source metadata on any image retrieved before the end of 2026. Once the migration completes, file identifiers and URLs will change, and some currently accessible images will temporarily disappear from search results as the new system indexes content. The city's digital affairs team has said a transition guide will be published before the rollout begins, though no specific publication date has been confirmed.