In Pompeii, archaeologists have employed artificial intelligence for the first time to reconstruct a victim’s image from the 79 C.E. volcanic eruption. This case highlights AI’s growing role in creating “historical” imagery, often without expert oversight. Roland Meyer, a media scholar at the University of Zurich, notes this trend reflects a deliberate effort to fill gaps in historical records through synthetic visuals.
Meyer warns that such AI-generated content risks deception. He explains: “These images don’t reveal the past but mirror our present expectations.” AI tools, trained on existing data, often reproduce clichés and biases, creating a distorted historical narrative.
Identifying Fake Historical Content
Quantifying AI-generated “historical” videos is challenging, but one immediate step is to question their plausibility. Ask: Could this scene realistically occur in the claimed era? Historian Sebastian Kubon emphasizes: “Videos from eras predating recording technology are inherently fictional, just as AI-generated ones are.”
Key red flags include anachronistic details, such as modern clothing in ancient settings, and unnatural movement patterns. The Pompeii video analyzed displays these traits, undermining its claim of authenticity.
Evaluating Visual and Source Credibility
A second approach involves examining visual details and provenance. For instance, a video falsely depicting U.S. soldiers commemorating Iwo Jima with a “US flag” in the background is clearly AI-generated. Obvious signs include glitching figures, inconsistent lighting, and watermarks indicating AI tools like Veo. The video’s description promotes “patriotic stories,” yet its reliance on a single iconic photo as a template exposes its lack of historical grounding.
Experts stress verifying sources. Authentic footage of the Iwo Jima moment does not exist, yet AI-recreated versions circulate. Similarly, a fake Chernobyl CCTV clip fails scrutiny when compared to archival records. Archives and experts remain critical for validation.
Cross-Referencing Verified Historical Records
Authentic materials include archival photos, academic publications, or museum collections. For example, the Iwo Jima photograph is preserved by the National Archives, while no verified video exists for Chernobyl’s disaster. AI-generated content bypasses these checks, prioritizing plausible visuals over factual accuracy.
This approach risks conflating speculation with history. As Mélanie Sengez, a historian, notes: “AI lacks understanding of context, producing appearances of truth through statistical patterns rather than evidence.” This undermines efforts to access genuine historical documentation, especially in periods without original recordings.
To counter this, researchers and the public must prioritize cross-referencing and critical analysis. While AI offers innovative tools, it cannot replace rigorous historical method.
