Paris Cleans Up Its Digital Streets: How the Capital Is Tackling Duplicate Images Online Compared to London, Berlin and New York

Paris city hall quietly launched an audit of its municipal image databases in March 2026, targeting thousands of duplicate and near-duplicate photographs cluttering the platforms that serve everything from tourism promotion to urban planning applications. The review, coordinated through the Direction de la Communication de la Ville de Paris, covers roughly 14 portals and sub-sites under the paris.fr umbrella — a sprawl that grew substantially after the city's digital push ahead of the 2024 Olympic Games.

The timing is not arbitrary. Post-Olympics legacy projects have left the city with an enormous volume of official photography: drone shots of the Seine-Saint-Denis velodrome, finish-line imagery from the Trocadéro, crowd stills from Pont d'Iéna. Much of that material was filed in haste across multiple servers by different contractors. Duplicates proliferated. Storage costs rose. And, more practically, public-facing pages began surfacing the wrong version of nearly identical images — a problem that undermined the city's efforts to present a coherent visual identity as it pivots toward Grand Paris Express corridor promotion and Seine waterfront regeneration campaigns.

What Paris Is Doing — and What It Is Not

The city is running a two-phase approach. The first phase, already under way at the Hôtel de Ville's digital services wing near Place de l'Hôtel-de-Ville in the 4th arrondissement, uses perceptual hashing — a technique that generates a fingerprint for each image and flags near-identical matches for human review. The second phase, planned for rollout in autumn 2026, would introduce machine-learning triage to rank duplicates by quality score before a human editor makes the final call on deletion or archiving.

That measured approach contrasts sharply with London's Government Digital Service, which by late 2025 had deployed a largely automated deduplication pipeline across gov.uk's image library, removing files with minimal manual oversight. Berlin's Senate Chancellery went further, contracting a private vendor in 2024 to run bulk deletion across the city's Landesarchiv-linked digital repositories — a move that drew criticism from archivists who argued that near-duplicate images sometimes carry distinct metadata value. New York City's Department of Citywide Administrative Services, meanwhile, still relies primarily on keyword tagging for image management, with no city-wide deduplication programme publicly announced as of mid-2026.

Paris, by contrast, is threading a line between the speed of London's automation and the caution of its own archivists at the Bibliothèque Historique de la Ville de Paris on Rue Pavée in the Marais, who have flagged concerns about permanent deletion of images that may have research or legal evidentiary value, even when visually redundant.

The Numbers Behind the Mess

French public administration digital storage costs are not uniformly published, but the Direction interministérielle du numérique — known as DINUM — indicated in its 2025 annual report that redundant data across French state platforms represented a measurable drag on cloud infrastructure budgets, which for Paris city-level operations ran to tens of millions of euros annually. The Olympic legacy archive alone is estimated internally to contain upwards of 400,000 images, with duplication rates in some event folders exceeding 30 percent, according to project documentation shared with municipal councillors in April 2026.

That figure tracks with patterns seen elsewhere. A 2025 study by the Reuters Institute at the University of Oxford found that large public-sector image repositories in European capitals typically carry duplication rates between 20 and 40 percent, driven largely by multi-contractor upload workflows and inadequate intake protocols.

For Parisians navigating the practical fallout — journalists pulling wrong images from official feeds, urban planners loading outdated aerial shots into planning software — the fixes cannot come fast enough. The city's autumn 2026 rollout of phase two will be the real test. If the machine-learning triage holds up on the Seine waterfront photography batch, which municipal project managers have identified as the most chaotic subset, the model could be extended to the Grand Paris Express corridor documentation archive by early 2027. If it does not, Paris may find itself borrowing a page from Berlin's experience: moving fast, deleting regrettably, and spending months rebuilding what was lost.