'It Erases Our Stories': SF Residents Speak Out on Duplicate Image Replacement in City Archives

The photographs are gone. Not lost in a fire or flood, but removed by software — algorithmic tools that flagged thousands of scanned images as duplicates and deleted them from San Francisco's public digital archive. Residents from the Tenderloin to the Excelsior are now pushing back, saying the city's rush to digitize and streamline historical records has cost them irreplaceable documentation of their own communities.

The issue surfaced publicly this spring when volunteers with the San Francisco History Center at the Main Library branch on Larkin Street noticed gaps in recently uploaded collections from the 1970s and 1980s. Images that archivists had catalogued were simply absent from the public-facing database. After internal review, library staff confirmed that an automated de-duplication process, part of a broader digitization push funded through the city's Department of Technology, had flagged visually similar images and retained only one version — deleting others without human review.

What Was Lost

For residents of the Western Addition, the stakes are personal. The neighborhood was razed and rebuilt through a federally backed urban renewal program in the 1960s and '70s, displacing thousands of Black San Franciscans. Photographs from that era are among the most sensitive and contested records in the city's holdings. Community members who rely on those images to document their family histories, support reparations advocacy, or simply bear witness say the automated deletions compound an old injury.

The problem is not unique to the Western Addition. At the Mission Cultural Center for Latino Arts on 24th Street, staff have spent months cross-checking their own donated photograph collections against what appears in the city's online portal. Several images documenting the Mission District's 1980s arts scene appear to have been collapsed into single representatives, with variant shots — different moments from the same event, sometimes showing different faces — removed. The Mission Economic Development Agency has separately flagged concerns to the city's Office of Civic Innovation about the long-term implications for community land trust documentation and neighborhood planning records.

Archivists point out that duplicate detection algorithms are built to match pixel similarity, not historical significance. Two photographs taken seconds apart at a community meeting in 1983 may look nearly identical to software, but each frame can contain different people, different expressions, different truths. The city's digitization contract, renewed in January 2026, covers an estimated 1.2 million documents and images across multiple departments. The de-duplication parameter was a cost-control feature, reducing cloud storage fees by an unspecified amount, according to public procurement documents reviewed by The Daily San Francisco.

Calls for a Moratorium

Advocacy groups are now asking the city to pause automated image deletion until a human-review protocol is in place. The San Francisco Public Library Commission is scheduled to take up the question at its July meeting, held at the Main Library's Koret Auditorium on Fulton Street. Community members plan to attend and deliver testimony.

The push comes as San Francisco's broader tech environment remains in flux. After years of layoffs, the city has seen an AI and automation boom that has accelerated the adoption of algorithmic tools across municipal services — from permit processing at the Department of Building Inspection to records management at multiple agencies. Critics argue that speed and efficiency metrics have overridden the kind of slow, careful judgment that archival work requires.

For anyone who believes their community's photographs may have been affected, the San Francisco History Center accepts inquiries by email and in person at the Larkin Street branch, Tuesday through Saturday. The center is also asking members of the public to submit original prints or negatives for comparison scanning — a process that can help staff identify what the algorithm removed. The library's digital services team has said it is auditing affected collections, though no completion date has been announced. Community members say that timeline is not good enough. Their histories, they argue, cannot wait for the next budget cycle.