Fact check: Spotting AI-generated 'historical' images

Archaeologists in Italy’s ancient city of Pompeii have employed artificial intelligence for the first time to reconstruct an image of a victim of the volcanic eruption that destroyed the region in 79 C.E. This marks a new phase in the use of AI to recreate historical moments, sparking debate among scholars about the implications of such technology in shaping public perception of the past. Experts caution that AI-generated “historical” images and videos are becoming increasingly common, often created without input from historians or archaeologists. Roland Meyer, a media scholar at the University of Zurich, argues that while AI allows for the recombination of historical imagery, it risks producing misleading representations of the past. He described the process as offering a “deceptive promise,” suggesting that AI-generated images reflect contemporary biases rather than accurate historical records. AI systems frequently produce errors, sometimes generating entirely fabricated content, referred to as “hallucinations.” These inaccuracies complicate efforts to distinguish real historical footage from AI-generated material. Historian and AI specialist Sebastian Kubon noted that many viral AI-generated “historical” videos blur the line between fact and fiction. A simple question, “Could this video have actually existed?”, often helps identify such content. For instance, a video purporting to depict Pompeii was easily dismissed due to its implausible nature, including movement resembling zombies and features inconsistent with ancient Roman aesthetics. Visual clues and careful examination of sources are essential in identifying AI-generated content. In one example, a video featuring U.S. soldiers holding the American flag showed unnatural movements and included a watermark indicating it was created using Veo, a generative AI tool. The video was based on a well-known 1945 photograph of soldiers raising the flag on Iwo Jima, yet no authentic footage of that exact moment exists. Experts emphasize the need to scrutinize both the content and context of such materials, questioning their origins, creators, and intended purposes. Critics argue that AI-generated images often replicate existing stereotypes and clichés about the past, reflecting current societal views rather than providing genuine insights into history. Jürgen Zimmerer, a professor of African history at the University of Hamburg, highlighted that visual sources have long been contentious, noting that authoritarian regimes historically manipulated images for propaganda. However, the widespread availability of AI tools has amplified the potential for misinformation on an unprecedented scale. Despite concerns, the integration of AI into historical research offers opportunities for innovation. Archaeologists and historians are exploring ways to use AI responsibly, ensuring that reconstructions are transparent and accompanied by clear explanations of their limitations. As the technology evolves, so must the standards for verifying and interpreting historical content. Scholars stress the importance of maintaining rigorous analytical practices to prevent the spread of misleading information. Efforts are underway to educate the public on recognizing AI-generated content through critical thinking and source evaluation. Educational initiatives aim to equip individuals with the skills needed to discern credible historical accounts from fabricated ones. As AI continues to influence how history is presented and understood, the role of experts in guiding public discourse becomes increasingly vital. Researchers are working to develop frameworks that balance technological advancement with ethical considerations, ensuring that the past is neither distorted nor misrepresented.