Their Photos Keep Disappearing Online. New Yorkers Say Duplicate Image Filters Are Erasing Their Stories.

Silvia Restrepo spent three years photographing every block party, quinceañera and street mural along Roosevelt Avenue in Jackson Heights. Last spring, she logged into the community archive she helps run for the Queens-based nonprofit Immigrants United Media Collective and found more than 400 images flagged and hidden — victims, she was told, of an automated duplicate-detection algorithm deployed by the platform hosting the archive. The photos were not duplicates. They were, in many cases, the only copies that existed.

The timing matters. With FIFA World Cup matches scheduled at MetLife Stadium beginning June 2026, city agencies and community groups have been racing to document New York's immigrant neighborhoods for cultural tourism initiatives, grant applications and historical preservation projects. Automated systems that remove or suppress images deemed visually similar are cutting into that work precisely when demand for authentic local documentation is highest.

A Tool Built for Spam Is Hitting Community Archives

Duplicate-image detection was designed primarily to fight spam and copyright infringement on large commercial platforms. The technology uses perceptual hashing — a process that assigns each image a numeric fingerprint based on pixel patterns — to flag near-identical files. When two images score above a similarity threshold, one is suppressed or deleted. The problem, according to digital archivists at the Metropolitan New York Library Council, headquartered on East 34th Street in Manhattan, is that community documentation often contains intentional repetition: multiple shots of the same storefront, the same protest corner, the same family at different events. Those patterns trip the filters.

The Bronx Documentary Center on Melrose Avenue has fielded complaints from at least a dozen local photographers and community historians over the past eight months, according to its public programming calendar and posted notices. Staff there have been advising contributors to rename files, alter metadata and manually stagger uploads — workarounds that add hours to what should be routine archival work.

For residents of Sunset Park, Brooklyn, the stakes are concrete. The Southwest Brooklyn Industrial Development Corporation has been compiling a visual record of the neighborhood's manufacturing corridor for a state economic development application. When platform filters flagged a batch of exterior photographs of factory buildings along 39th Street as duplicates — the images were shot at different times of day and in different seasons — the organization had to resubmit its documentation package, delaying the process by roughly six weeks.

Who Gets to Decide What's a Duplicate?

The people most affected are not coders or policy analysts. They are residents who showed up with phones and point-and-shoot cameras to record something they feared would otherwise go unrecorded. In Flushing, Queens, volunteers with the Flushing Remembrance Project — a grassroots group that has been cataloguing the neighborhood's rapid commercial redevelopment since 2019 — say they have lost access to nearly 200 images of storefronts on Main Street that no longer exist. The buildings came down. The images got flagged. There is no appeal process that community members can easily navigate.

The financial dimension is not trivial. Small nonprofits operating on grants under $50,000 a year rarely have the technical staff to challenge automated moderation decisions or migrate archives to more permissive platforms. Migrating a mid-size photo archive — roughly 5,000 images — to a self-hosted solution can cost between $3,000 and $8,000 when factoring in storage, labor and ongoing maintenance, according to pricing guides published by the Digital Preservation Coalition.

City Council's Committee on Technology has not yet scheduled hearings specifically on automated image moderation, though the broader question of algorithmic accountability in public-facing platforms has been part of the local policy conversation since the passage of Local Law 144 of 2021, which addressed automated employment decision tools. Advocates say the documentation problem deserves similar scrutiny.

For now, groups like Immigrants United Media Collective are doing what New York community organizations have always done when formal systems fail them: improvising. They are building mirror archives on hard drives, cross-posting to multiple platforms, and teaching members to export full-resolution files before uploading anything to the cloud. The Bronx Documentary Center is planning a free workshop in August on archival best practices for community photographers. Registration details are expected on its website by mid-July. It is not the systemic fix anyone wanted, but it is a start.