Boston Facebook Page's AI ‘History’ Videos Spark Backlash From 14,000 Followers

A Facebook page called Historic Boston Archives has built a following of 14,000 people by posting videos that claim to bring old Boston photographs to life — except the videos are riddled with fantasy cathedrals, driverless cars, and signs with garbled spelling that give away their artificial intelligence origins. The page takes real, colorized historic photos and runs them through AI animation tools, and its own followers have started calling out the results as fake.

As reported by Boston.com, the page's photos generally appear authentic, since they are often colorized real images, but the source institutions are never credited. It is the videos generated from those photos that have raised alarms, with the outlet's reporting describing them as AI hallucinations rather than accurate depictions of Boston's past. Separately, Universal Hub reported this week that the page's AI-generated clips included a massive fictional cathedral looming behind Trinity Church in a Copley Square video and a modern Amazon delivery box appearing in a purported 1970s scene of Boston Common.

Spelling Errors and Driverless Cars Give the Fakes Away

One video purporting to show the Custom House in a 1970s scene misspelled the word Boston on a sign and showed cars driving in the wrong direction, per the same Boston.com report. A Copley Square video set in 1920 showed cars being driven with nobody behind the wheel. Followers identified an Oak Square video, set in 1970s Brighton, as fake because of nonsensical signage in the frame.

Another clip depicted Guay's Bakery in Dorchester in a purported 1930 scene, layering old-time music and a bizarre voiceover over the footage, complete with invented dialogue for AI-generated men standing in the rain. Followers of the page have increasingly used comment sections to flag these details, criticizing the videos outright as AI slop.

Why the Technology Keeps Getting It Wrong

John Wihbey, a Northeastern University professor and assistant vice chancellor for strategic AI initiatives, told Boston.com that these kinds of errors are common in current AI video generation and are likely to improve as the technology evolves. AI systems still struggle to accurately render signage, lettering, symbols, and vehicle direction, which is why typos and cars without drivers keep surfacing in the Historic Boston Archives clips.

Wihbey drew a distinction between problematic content and legitimate uses of the technology. He said well-documented, clearly labeled AI content made for a pro-social purpose does not qualify as AI slop, but content made carelessly for engagement, monetization, or fun is more properly considered exactly that. The page's owner did not respond to Boston.com's request for comment.

The National ‘Slop’ Problem Behind the Local Videos

The term has entered the mainstream lexicon: Merriam-Webster selected “slop” as its 2025 Word of the Year, defining it as low-quality digital content produced in bulk via generative AI to capture attention and farm engagement, according to Vervocity. A study in the Harvard Kennedy School Misinformation Review, examined via PMC, found that page operators create low-effort AI images and videos specifically to build large followings they can later monetize through platform payouts, ad revenue, or merchandise.

That financial incentive helps explain why pages avoid labeling their content, even as Meta has moved to require it. In July 2026, Meta updated enforcement across Facebook and Instagram to automatically append an AI info disclosure tag to posts when metadata or automated classifiers detect generative AI editing, according to Marketing Dive. But industry analytics reported by Sybrid show that an explicit AI tag can cut engagement by 15% to 30%, giving page owners a clear financial reason to skip voluntary disclosure.

What's at Stake for Public Memory

The risk goes beyond annoyance. Visual images shape how people remember things, and researchers have warned that AI historical videos risk interfering with humans' accurate memories of the past, per Boston.com's reporting on Wihbey's comments. The American Psychological Association warned in a July 2026 publication that cognitive offloading to AI systems and exposure to inaccurate synthetic media can impair working memory and make viewers prone to accepting false details as real memory.

Compounding the problem, visual cognition research published in Interalia Magazine found that human viewers struggle to distinguish authentic photographs from synthetic AI imagery, with detection accuracy hovering near chance levels around 50%. That helps explain why casual scrollers might not catch the Historic Boston Archives errors until someone points out the fake cathedral or the misspelled street sign.

A Different Approach to AI and Boston's Archives

The contrast with institutional efforts is notable. The Boston Public Library launched an initiative in March 2025 with Harvard Law School's Institutional Data Initiative to use AI for digitizing, indexing, and expanding public access to hundreds of thousands of authentic historical documents and photos — using the technology for metadata organization and optical character recognition rather than generative video animation. Historians and archivists have separately warned, as reported by Hürriyet Daily News, that a flood of AI-generated fake history content on social platforms risks clouding public understanding of documented events, particularly for eras with limited surviving film documentation.