Fake Bezos and Statham Hair Photos Debunked by Finding the Originals
The Verge reported on November 3, 2026 that viral images showing Jeff Bezos and Jason Statham with far more hair than their usual shaved look are fake photo edits.
The images circulated on X, Threads and Bluesky. In testing the fakes, The Verge found that none carried Content Credentials or SynthID markers, so it could not confirm that AI tools were used to make them. The outlet also could not fully establish when the images were first posted or who created them.
The reverse-image search is where the story was resolved. It surfaced unaltered photographs of Bezos taken against the same background and wearing the same outfit, with no added hair and a watermarked image credit. Those originals date to 2023. For Statham, the search found watermarked versions of the original photographs, taken by a paparazzo in 2011.
Each fake therefore has an authentic counterpart shot in the same setting. That pairing points to edits made on top of existing photographs. The matching backgrounds and clothing support that conclusion, though the method used to alter the images remains unidentified.
In my view, the episode is a useful reminder that provenance tooling is only one layer of verification. Content Credentials are designed to attach signed metadata to an image at capture or edit, and SynthID is an invisible watermarking scheme. Both depend on the file arriving intact. Screenshots, re-encoding and re-uploads can strip that metadata, and an image without markers is not thereby proven authentic or synthetic. The missing markers told reporters little on their own.
The broader context here is that the more durable check was the most old-fashioned one: finding the original. A reverse-image lookup, a watermark credit and a date on an earlier photograph formed a verifiable trail that did not depend on any detector's verdict. The method needs no specialised tooling, and an ordinary reader on a phone can run it in under a minute.
The unresolved parts are real. Nobody has been identified as the author of the edits, and the posting history of the images could not be fully traced. Those gaps matter for anyone trying to assign responsibility or to study how such fabrications spread across X, Threads and Bluesky. The episode does not settle them.
What the case does show is that a convincing edit of a public figure can be checked against a public archive of originals. For a feed moving at speed, that remains a practical defence, and it does not require trusting a model's answer about whether a picture was made by machine.
