'My Grandmother's Face Was Replaced by a Stranger': SF Residents Speak Out on Duplicate Image Removal

Dozens of San Francisco residents say they have lost irreplaceable personal photographs after automated duplicate-image detection systems purged their uploads from shared community archives, neighborhood Facebook groups, and city-affiliated digital history platforms — sometimes deleting originals rather than copies.

The problem surfaced publicly in late June when the Western Neighborhoods Project, a nonprofit that maintains a digital archive of San Francisco history, acknowledged it had experienced a batch-processing error tied to a third-party deduplication tool. Members who had contributed scanned family photos stretching back to the 1940s reported finding their submissions gone, replaced in some cases by placeholder thumbnails or, more jarringly, by unrelated images pulled from similar metadata tags.

The Human Cost on Balboa Street and Beyond

For families in the Outer Richmond and the Sunset — neighborhoods with deep Chinese American, Russian American, and Irish American roots — the deletions hit harder than a server error. One Balboa Street resident described watching a digitized photo of her grandmother, taken outside a Clement Street fish market sometime in the 1960s, vanish from the Richmond District History Archive's shared folder and be replaced by a stock image of an unidentified woman in a similar coat. She has been trying to recover the original scan for three weeks.

The Richmond District History Archive, which operates out of the Park Branch Library on Arguello Boulevard, relies on volunteer-managed cloud storage and has no dedicated IT staff. A volunteer coordinator there said the archive's deduplication software — licensed through a subscription service that auto-renewed in January 2026 — flagged roughly 340 images in a single overnight run, deleting what it classified as lower-resolution duplicates. In several documented cases, the "duplicate" was the only surviving copy.

Similar complaints have emerged from users of Cuentos SF, a Mission District oral history and photo documentation project that works with Latino families in the 24th Street corridor. Participants there had submitted scanned prints as part of a 2025 neighborhood memory initiative; some now report their contributions are either missing or swapped with mismatched images from other contributors' folders.

Why It's Happening Now

The timing is not coincidental. Across the tech sector, AI-assisted image processing tools have dropped sharply in price over the past 18 months, making sophisticated deduplication software accessible to nonprofits and small civic organizations that previously could not afford it. The same compression in cost that opened the door to these tools also removed the friction that once forced administrators to review deletions manually before executing them.

The Internet Archive, which hosts a San Francisco-specific collection at its Funston Avenue headquarters in the Richmond District, has logged a noticeable uptick in recovery requests from Bay Area users since April 2026. The organization does not publish granular city-by-city request data, but staff there have described the trend publicly in blog posts without citing specific numbers.

For everyday residents, the options are limited and often expensive. Professional photo recovery services in the SoMa district quote between $150 and $600 per session for digital file recovery, depending on the storage format and extent of the loss. Cloud platforms typically retain deleted files in recoverable form for 30 days — after that, they're gone.

The Western Neighborhoods Project has said it is auditing the full scope of the deletion event and will publish results by the end of July. Anyone who contributed images to the archive before June 15 is urged to check their submission confirmation emails and compare file names against what remains accessible in the public gallery at OpenSFHistory.org. The Park Branch Library on Arguello is hosting a drop-in recovery session on July 12 where volunteers will help affected contributors cross-reference their original scan files. Cuentos SF is asking participants to contact the project directly through its 24th Street office before attempting any third-party recovery service, to avoid compounding the data loss.