The Fabrication of Heroism
Welles Crowther died on the 104th floor of the South Tower during the September 11, 2001, terrorist attacks. He was a 24-year-old equities trader and volunteer firefighter who led survivors to the only functioning stairwell. His actions saved at least a dozen lives. Despite the historical record of his sacrifice, a synthetic image of Crowther rescuing victims recently went viral. This artifact represents a new, dangerous trend in how digital tools overwrite history.
There are no authentic photographs of Crowther in action during that day. The image currently spreading across social media is a modern construction produced by generative artificial intelligence. It depicts a man in a button-up shirt and tie, a visual narrative that does not align with eyewitness accounts or historical archives. This fake image serves as a warning that our collective memory of the September 11 attacks is now under threat from fabricated visuals that look indistinguishable from genuine documentation.
The Failure of Detection Tools
When we tested this viral image against major AI platforms, the results were alarming. ChatGPT examined the file and claimed it was a genuine historical photograph. It insisted the image possessed features expected from news photos of the era. The software failed to identify the source as synthetic, proving that these tools are not equipped to police their own output.
Other platforms fared no better. Grok initially labeled the image as a well-known historical record before backtracking under pressure. Claude asserted that the image was a dramatization from a documentary, which was also false. When pressed further, these systems continued to offer hallucinations. They incorrectly cited irrelevant miniseries or claimed they could not detect watermarks like Google’s SynthID. These failures show that users cannot rely on AI to verify the truth of what they see online.
Historical Erasure Through Speed
The ease of creating photorealistic fakes has destroyed the friction that once limited digital misinformation. During the 2010s, creating a deceptive image required skill with software like Photoshop. Now, any user can generate a convincing lie with a few text prompts in seconds. This speed allows false narratives to travel faster than truth, polluting historical databases and social feeds alike.
Roughly 25 percent of Americans in 2026 were not alive on September 11, 2001. For these younger citizens, the visual record on the internet becomes their primary source of understanding. When that record is flooded with AI-generated slop, the reality of the day fades. The danger is not just that people believe the fake images; it is that the abundance of AI content leads people to doubt reality itself. Public figures have already started dismissing genuine evidence as fake simply because they can blame AI, further muddying the record for everyone. We must prioritize primary sources and documented history, or we risk losing the truth to the convenience of an algorithm.