I remember the first time I saw that red bandana: a headline, a face, a story stitched into September 11th. You scroll, you stop, you feel steadier for a moment—until the image mutates into something that wasn’t there. That slow, small corruption is how memory can leak out of history.
I’ve chased false photos before. You should know how fast they spread now, how easily trusted systems hand you a lie. I’ll walk you through the mess: why an AI image of Welles Crowther—the Man in the Red Bandana—went viral, why smart chatbots defended it, and what it means when the internet rewrites a day millions still count by.
On a viral image that appeared overnight
The image claimed to show Welles Crowther rescuing people on 9/11, and it arrived like a cold second wave.
You’ve already heard Crowther’s story: a 24-year-old equities trader and volunteer firefighter on the 104th floor of the South Tower, credited by multiple witnesses with guiding and carrying people toward the single working stairwell before dying when the tower collapsed. That heroism is real; the photo is not.
On what AI did to that story in minutes
The fake image surfaced on social feeds and spread faster than any one person could follow.
I ran it through the usual checks and watched chatbots I expect to be cautious bend under pressure. ChatGPT initially responded that the image “very likely is not AI-generated” and called it consistent with news photography. Grok first called it a known 9/11 photo before admitting it might be fake. Claude insisted it looked like a dramatization and then provided a reverse-search claim tying it to a documentary. Gemini flubbed checking SynthID and then reversed course. These are not distant mistakes; they’re the lived products of systems trained to fill gaps with confidence.
Trump on a deceased 9/11 rescuer:
“The Man in the Red Bandana. Have you heard of him? He’s become more famous than me. I don’t like it. Very unhappy. But he’s become much more famous than me.”
— FactPost (@factpostnews.bsky.social) September 8, 2026 at 8:25 AM
On how trusted brands messed up the check
A reporter ran the image through ChatGPT, Grok, Claude, and Gemini and got four different narratives.
That’s the point: platforms and their models—OpenAI’s ChatGPT, xAI’s Grok, Anthropic’s Claude, Google’s Gemini—have become front-line arbiters of authenticity, and they can be wrong in ways that amplify falsehood. Google’s own SynthID, a watermark system for images created with Google tools, was misreported by Gemini; Claude insisted on documentary origins; Grok flipped after the reporter pressed it. The systems are useful, but their confidence can be mistaken and contagious.
Is the viral “Man in the Red Bandana” image real?
No. There are no confirmed contemporaneous photos of Crowther in action on 9/11. Reverse-image searches show the viral photo had no prior publication history. The scene in the 2017 documentary Man In Red Bandana features a reenactor in a white T-shirt—different from the tie-and-button-up shirt seen in the AI image—and that film clip does not match the viral frame.
On how images shape generational memory
About 25% of Americans alive in 2026 were born after 2001, and many never saw the day firsthand.
When photographs are faked, the archive that future generations inherit changes. AI slop is a virus in the bloodstream of memory. These fakes aren’t neutral; they add false texture to the day, shifting how people who never lived it imagine it happened.
Can AI image detectors be trusted?
Not completely. Detection tools exist—Google’s SynthID, reverse-image search engines, forensic suites—but models hallucinate, detectors have false negatives, and even journalists found ChatGPT and Gemini confidently wrong. Relying on a single verdict from an AI system is a mistake; cross-checking with original sources, eyewitness testimony, and archival footage remains essential.
On platforms and the incentives that warp truth
The platforms reward speed, clarity, and emotion—often over accuracy.
Facebook, X, Google, Tubi, mainstream outlets, and fringe pages all play a role. A sensational image earns engagement, and engagement overrides patience. Donald Trump’s public comments praising Crowther and then pivoting the spotlight toward himself are an example of how political figures can twist attention; the result is attention that flattens complexity into sound bites.
How will AI images change how we remember historic events?
If we let feeds become the archive, history will look curated by whoever controls the prompts. The funhouse mirror for history is already in operation: images that never existed carry emotional authority because they look real, are shared by trusted accounts, or are affirmed by overconfident AIs.
On what you can actually do when you encounter a suspect image
When you see a historic photo that feels definitive, take three simple steps.
I recommend: 1) do a reverse-image search, 2) check multiple AI detectors and human sources, and 3) look for contemporaneous provenance—dates, news wires, archival footage. Treat a single AI verdict as a prompt, not proof. Tools help—TinEye, Google Images, SynthID checks inside Gemini—but your skepticism is the best filter.
We used to fight Photoshop fakes that required skill and time; now anyone can generate photorealistic frames in seconds. That lowers the bar for deceit and raises the cost of trust.
I want you to keep Crowther’s actual story intact: witness accounts credit him with saving at least a dozen people, and President Donald Trump awarded him the Presidential Medal of Freedom—an honor that reopened attention to a real life, not a manufactured scene.
We are at a crossroads where archives and feeds can no longer be treated as the same thing. If images become the raw material of memory, who gets to tell the story—and what will you believe when the next fake hero appears with a compelling bandana?