Nobody Ever Photographed The Man In The Red Bandana. ChatGPT Called This Picture Of Him Genuine.
No photograph of Welles Crowther in the South Tower exists. Three days after his family accepted his Medal of Freedom, a fake one went viral, and the four biggest chatbots could not say so.
The Tribute in Light, New York, September 2026. Photo: Ponor, CC BY SA 4.0, via Wikimedia Commons.
On Tuesday, September 8, 2026, Alison Crowther and her daughter Paige stood on the Ellipse in Washington and accepted the Presidential Medal of Freedom for a son and brother who has been gone for twenty five years. Welles Crowther was 24, an equities trader and a volunteer firefighter, and survivors of the South Tower remember him only as a voice and a red bandana tied across a face in the smoke. CBS News credits him with helping as many as 18 people out.
There is no photograph of him doing it. Not one. Everything the world knows about his last hour comes from the people who made it out.
By the anniversary itself, three days after the ceremony, a photograph of him doing it had spread across Facebook. And when a reporter asked the most popular AI chatbots on earth whether it was real, ChatGPT said it looked like a genuine historical photograph.
The picture that should not exist
Matt Novak at Gizmodo published the story at 3:20 p.m. Eastern on September 11, 2026, under a headline that says everything: “Never Forget, AI Slop Is Polluting 9/11 History on the Internet as We Speak.”
The image shows the Man in the Red Bandana apparently carrying people to safety inside the World Trade Center. It is emotional, well lit and completely invented. Reverse image searches turned up no trace of it before that week, which is exactly what you would expect from a picture of a moment nobody ever photographed. Novak also noted it was not the only one: other AI images of Crowther had already been circulating this year.
The closest thing to footage is a reenactment. The 2017 documentary Man in Red Bandana used an actor to very briefly recreate some of what Crowther did, and that actor wore a plain white T shirt. The viral image puts him in a button up shirt and tie. It is not even a copy of the dramatization. It is a new memory, manufactured from nothing.
Four chatbots, four answers, one truth nobody landed on
Novak did what millions of people now do when a picture looks too good: he asked the machines. Here is what he got, as reported by Gizmodo.
- ChatGPT said the image very likely was not made with AI and appeared to be “a genuine historical photograph.”
- Grok first called it a well known photograph from the attacks, and only admitted it might be fake after being pushed.
- Claude said it was “a reenactment/dramatization photo” and “not created with AI.”
- Gemini initially refused to check the image for Google’s own SynthID watermark, then contradicted itself once it was challenged.
Read that list again with the facts beside it. There is no well known photograph. There is no reenactment still that matches. The only honest answer available was “no photo of this moment has ever existed,” and not one of the four led with it.
Why this one hurts more than a fake celebrity picture
An AI chatbot vouching for a fake picture of a pop star is embarrassing. Vouching for a fake picture of a dead man’s final act of courage is something else, because the audience for 9/11 is changing.
Gizmodo, citing U.S. News & World Report, put a number on it: roughly 25% of Americans alive in 2026 were not born yet on September 11, 2001. A quarter of the country learns this day secondhand. For them there is no memory to check a picture against, only a search box, and increasingly a chatbot sitting on top of that search box ready to say “genuine.”
Think about who that picture reaches. A high school kid who just watched the news clip from the Ellipse types the name into a phone. Up comes a heroic image. They ask the assistant built into the app whether it is real. The assistant says yes. That kid now carries a picture of an event that never happened, of a real man, and will pass it on with total confidence, because they checked.
That is how you corrupt history. Not with one big lie. With a million small confirmations from a tool people have been told is smart.
The family did not need this
For twenty five years the story of Welles Crowther has been told as a story about what one person can do in the worst hour of his life. The story works because it is true and because it is built from testimony, from survivors who described being led through smoke and flames by a young man with a red bandana across his face.
A fabricated photograph does not add to that. It cheapens it. It invites every skeptic who ever wanted to wave away a hero story to say the evidence was faked, and the moment someone finds out the famous photo was AI, the doubt splashes onto everything else. The real testimony. The real medal. The real mother holding it.
What should have happened
This is not a hard case for a machine that claims to reason. The name in the question is famous. The absence of any photograph of him in action is part of his story, repeated in almost every account of it. A tool that knows who Welles Crowther was should know that, and should say it before it says anything about lighting or grain.
Instead the four biggest assistants did what they keep doing on this site’s pages: they produced a confident answer shaped like the question. “Is this real?” gets “it appears genuine,” because that is what an answer looks like.
If you see the red bandana picture this week, do the thing the chatbots would not. Say out loud that no one photographed him in that tower. Then tell the true part, which needs no help: a 24 year old with a bandana over his face led strangers through the smoke, and as many as 18 people are credited to him.
