AI-Generated Holocaust Images Fuel Historical Skepticism, Experts Warn

The Auschwitz Memorial has been documenting a quiet crisis unfolding across Facebook: accounts spreading images of the Holocaust that never existed, conjured entirely by artificial intelligence. A mother saying goodbye to her son before separation at Auschwitz. A couple holding hands one final time in a concentration camp. A Jewish man saluting the American flag after liberation. The narratives are always emotionally pitched, the photographs convincing enough at first glance—but they exist nowhere in any archive, only on pages of dubious credibility, bearing the telltale marks of machine generation. The memorial's statement was direct: by allowing these distortions to circulate and gain visibility, Meta contributes directly to the erosion of factual understanding about Auschwitz's complex history.

What makes this problem particularly acute is how easily it spreads. Nuno Moniz, a machine learning researcher at Notre-Dame University, explains that these fabrications move faster through social networks than authentic historical documents. The reason is algorithmic: emotional content simplifies the message, strips away context and nuance, and algorithms reward exactly that kind of superficial engagement. The images are designed not to document history but to generate clicks. A BBC investigation revealed that some accounts monetize this content, with one page manager claiming he could earn a thousand dollars monthly with 300,000 followers and "premium" content for Western audiences—because, he noted, Western views are worth eight times more than Asian ones in the system's calculation. On platforms like TikTok and YouTube, where any user can monetize video, AI-generated content proliferates even more freely. According to a Kapwing report, more than twenty percent of YouTube Shorts shown to new users consists of what researchers call "AI slop," low-to-medium-quality machine-generated material.

The deeper danger, though, is not the monetization itself but what it does to historical consciousness. Miguel Cardina, a historian and researcher at Coimbra University's Centre for Social Studies, identifies a cascading skepticism: if AI now produces so many false images, how can anyone be certain that an authentic photograph is actually authentic? Pushed to its logical extreme, this reasoning enables the denial of any historical moment whatsoever. The past risks becoming, as Cardina puts it, "a form of entertainment dictated by interactions"—stripped of reality, reduced to aesthetic creation rather than historical documentation. This echoes an older tactic: Stalin erased Trotsky and other "enemies" from photographs, but he needed specialists to do it. Today, anyone with access to an AI tool can accomplish the same erasure at scale.

Nor is this merely a problem of individual deception. Inês Narciso, a researcher at ISCTE and the CyberPeace Institute, points out that even unprofitable pages hold value. A political campaign manager might pay thousands of euros to acquire a page with 200,000 followers. The barrier to building large audiences has collapsed: you no longer need communication skills, just an idea and a machine to execute it. Narciso invokes a term coined by writer Cory Doctorow—"enshittification"—to describe how internet quality degrades over time while users remain trapped by platform dependency. She warns of a plausible scenario within five to ten years: the internet becomes nothing but "garbage, photocopies of photocopies," with AI systems themselves training on fabricated sources and producing increasingly strange amalgamations of artificial creation.

Regulating AI itself is nearly impossible, Moniz acknowledges. These are machines without agency, without reflection, easily manipulated—a structural characteristic unlikely to be solved. The responsibility must fall to the platforms where images are shared. Yet among dozens of pages identified by Público, only one labeled content as AI-generated, using Meta's watermark. The rest offered no indication whatsoever. The Auschwitz Memorial has reported multiple pages spreading false Holocaust content to Meta; the accounts remain active. Without platform accountability, the burden shifts elsewhere: to visual literacy in schools, to academic and journalistic work. Cardina emphasizes that teaching historical literacy must begin early, in primary school, and continue through university—students must learn to distinguish documents, critique sources, recognize manipulation, understand who created an image and why.

The scale of the problem is staggering. More than a billion people sit in front of screens searching for information about conflicts and receive countless variations of the same story, each stripped of depth and complexity. If social media becomes the primary source of historical knowledge, if there is no grounding in rigorous historiography and no visual literacy to detect fabrication, the probability of believing any convincing falsehood becomes very high. History education, then, becomes not a luxury but a defense: the more fragile and superficial our knowledge of the past, the more vulnerable we are to accepting entirely manipulated images as truth.