Does a video on social media show what posts described as Iran's Islamic Revolutionary Guard Corps, "hiding in schools thinking Israel won't bomb them"? No, that's not true: It's not real footage. The clip was made with generative AI from a photo that predated the Feb. 28, 2026, U.S.-Israel attack on Iran, as well as the following escalation.
The claim appeared in a post (archived here and here) published on X on March 15, 2026. It opened:
IRGC Hiding in schools thinking Israel won't bomb them. Doesn't matter, you'll still be killed.
This is what the image attached to the post looked like on X at the time of writing:
(Image source: screenshot of post by @IranSpec on X.)
Reverse image searches did not produce a match with an earlier video file, but helped to find several examples of the corresponding still image here (archived here) and here (archived here). Those pages went online on Jan. 8, 2026, and that meant that the picture couldn't show the war that followed the Feb. 28, 2026, U.S.-Israel attack on Iran.
When Lead Stories searched a fragment of the original image -- the unique front gate -- Google's "About this image" tab was able to recognize its contents and said that the image was "at least 2 months old":
(Image source: Google.)
Lead Stories tested the video from the post with the AI-generated content detection tool of hivemoderation.com. According to it, the clip was 82.6% likely to contain AI:
(Image source: Hive Moderation.)
The two most recent models from 2025 available in DeepFake-o-meter, another AI detection tool, showed that the probability of the video being AI-generated ranged from 37.5% to 100%:
(Image source: DeepFake-o-meter.)
A comparison between the still image and the video revealed discrepancies, suggesting the photo had been converted to a video using generative AI.
For example, we can see how the front gate's outline becomes distorted for no reason, even before the time the smoke enters the frame for the first time (click to view larger):
(Image sources: X, InVid.)
Additionally, what looks like shadows on the ground in the still image somehow become broken wires in the video:
(Image sources: X, InVid.)
OSINT analyst and media literacy educator Tal Hagin, who was the first to suggest that the photo was "motionized" by AI in this case (archived here), wrote that "all of the people are in the exact same positions as the old image" and that "AI messed up the front gate, with it adding/removing features not present in the real school gate".
