Grok told users a real Gaza photo was from 2014 Syria. Then it told them a real malnourished girl was from Yemen in 2018.

Review

In July 2025, a photo of a girl seeking food at a community kitchen in Gaza went viral. It was taken by AP photographer Abdel Kareem Hana on July 26, 2025. When X users asked Grok, the platform’s AI chatbot, to verify the image, Grok told them it showed a Yazidi girl fleeing ISIS in Syria in 2014. The false response spread rapidly. Users began repeating Grok’s claim and calling for the authentic AP image to be flagged as misinformation.

In August 2025, AFP photojournalist Omar al-Qattaa published an image of a severely malnourished nine-year-old girl, Mariam Dawwas, in Gaza City. When users asked Grok to verify this photo, the chatbot identified it as an image of a Yemeni child from 2018. The false response spread widely, and a French lawmaker was publicly accused of spreading disinformation simply for having shared the authentic AFP image.

When challenged, Grok insisted: “I do not spread fake news; I base my answers on verified sources.”

The Iran war: Grok as misinformation engine

During the February-March 2026 US-Israel war with Iran, Elon Musk encouraged users to “fact check and ask questions about any post” using Grok. The results were catastrophic:

• A video of a fire at Glasgow’s Central Station was shared as “firefighting efforts in Tel Aviv.” When users asked Grok to verify, it confirmed the fire was in Tel Aviv, consistent with Iranian missile impacts. X’s own Community Notes had already identified it as Glasgow.

• A real video of fires in Tehran’s drainage canals went viral. Grok confidently told users it was from a 2017 wildfire in California — the Skirball Fire on LA’s I-405 freeway. When users provided comparison footage and Community Notes corrected the record, Grok doubled down, insisting the clip still matched the 2017 wildfire.

• When asked to verify footage of a strike on a girls’ school in Minab, Iran, Grok claimed the video showed an ISIS attack on a school in Kabul in 2021. Subsequent analysis by the New York Times, Reuters, and Bellingcat confirmed the school was hit by US missiles. At least 160 people were killed, the vast majority young children.

• Grok falsely claimed a real video from Tehran showed Italian tricolor flags on a European-style highway. It produced AI-generated images of destruction, adding to the “AI slop” on the platform.

The liar’s dividend, automated

This is the liar’s dividend operating at machine speed. When an AI chatbot embedded in a social platform tells users that a real photo is fake, it does not just make one error. It trains the audience to distrust real evidence. And when the platform owner tells users to use that chatbot for fact-checking, the authority of the platform is lent to the error.

The Grok cases are not hallucinations in the technical sense — they are the chatbot confidently generating false provenance for real images, and then refusing to correct itself when challenged. The result is a system that takes real documentation of human suffering and rewrites it as something else, while the users who shared the real image are accused of spreading fake news. That is not a bug. It is a propaganda function.

Verdict: False. Grok’s claims were wrong in every documented case. The real photos were real. The AI chatbot was the source of the misinformation, and the platform owner directed users to trust it.