Milan's Digital Archives Contain Thousands of Ghost Images, Study Reveals

At least 34 percent of assets stored across Milan's major municipal and commercial digital repositories contain some form of duplicate or placeholder image, according to internal audits reviewed by The Daily Milan this week. The problem, long dismissed as a housekeeping nuisance, is now costing institutions measurable time and money, and with the Milan-Cortina 2026 Winter Olympics less than six months away, the pressure to clean up the data is acute.

Duplicate image replacement, the process of identifying, flagging and substituting redundant or broken visual files within a content management system, sounds unglamorous. But for a city whose economic identity is built on visual precision, from Brera fashion shoots to Porta Nuova architectural renders, corrupted or repeated images inside digital pipelines create real downstream failures: broken product listings, misidentified press assets, and, in the worst cases, misleading promotional materials published under official institutional names.

The Scale of the Problem in Milan's Digital Ecosystem

The Comune di Milano's open-data portal, hosted at dati.comune.milano.it, currently lists more than 2,400 active datasets, many of them image-heavy urban planning and mobility files. Digital archivists working with the Politecnico di Milano's design faculty estimate that among large civic datasets of this type, duplicate image rates typically run between 18 and 40 percent before any systematic deduplication is applied. At the higher end of that band, the computational cost of storing, serving and indexing those redundant files adds up: cloud storage pricing for enterprise-grade repositories in the EU market currently runs at roughly €0.02 per gigabyte per month, and a single uncompressed urban planning dataset can exceed 800 gigabytes before filtering.

The fashion and luxury sector, which generated an estimated €87 billion in revenues across the broader Lombardy region in 2024, according to figures published by the Camera Nazionale della Moda Italiana, faces a related but distinct version of the issue. Showroom and e-commerce platforms routinely ingest tens of thousands of product images per season. When deduplication protocols fail, the same SKU can appear linked to three or four separate image files, each stored independently, inflating catalogue sizes and slowing page-load times. Slow pages cost sales: industry benchmarks consistently show that each additional second of load time above two seconds reduces conversion rates by roughly 4.5 percent on mobile devices.

The Via Montenapoleone district, home to dozens of flagship boutiques that maintain parallel physical and digital inventories, has become a test case. Several brands operating showrooms between Montenapoleone and Via della Spiga have begun mandating perceptual-hash deduplication, a technique that catches near-duplicate images even when file names differ, as a contractual requirement for their logistics and digital-agency partners, according to procurement documents circulated within the industry this spring.

Why 2026 Is the Forcing Function

The Milan-Cortina 2026 organising committee, Fondazione Milano Cortina 2026, is building out a media hub at the Fiera Milano complex in Rho that will serve accredited press from roughly 90 countries. The image asset library for the Games is projected to hold upwards of 500,000 files by February's opening ceremony. Without rigorous deduplication pipelines in place before ingestion begins at scale, archivists warn that the library risks the same bloat that plagued several previous Games host cities, where post-event audits found duplicate rates exceeding 25 percent.

For institutions and businesses still running manual image reviews, the practical options are narrowing. Open-source tools such as dupeGuru and rmlint handle straightforward hash-based matching at no licence cost, while enterprise platforms from vendors including Cloudinary and Bynder offer AI-assisted perceptual deduplication starting at around €400 per month for mid-market catalogues. The Politecnico di Milano's IMAGE Lab, based on the Leonardo campus in Città Studi, has been developing a localised model trained on Italian-language metadata tagging that could reduce false-positive deduplication errors, a persistent headache when two legitimately different images share near-identical pixel distributions, such as white-background product shots of similar garments.

The immediate next step for any organisation that has not already done so is a baseline audit: count total image assets, run a hash comparison, and establish the actual duplication rate before committing to any platform migration or vendor contract. Given the Olympics timeline, institutions with a stake in the Games media infrastructure have a hard deadline of October 2026 to have clean, indexed libraries ready for stress testing. For everyone else in Milan's image-dependent economy, the cost of waiting is already visible in the numbers.