The graves, freshly dug, lie in neat rows of 20 throughout. More than 60 have already been carved out of the earth, with a couple of clusters of folks standing gathered around them. Dozens extra are marked out on the floor in entrance: small chalk rectangles, with diggers poised to finish their job.
The cemetery of Minab, photographed as it prepares to bury greater than 100 of the city’s younger ladies, is one of the defining photographs of the US-Israeli warfare on Iran, bluntly capturing the devastating civilian toll.
But is it actual?
Ask Gemini, the AI service powered by Google, and the reply you obtain is not any – in reality, Gemini claims the {photograph} is from two years earlier and greater than 2,000km (1,240 miles) away. Rather than graves for small ladies killed by a missile, the picture “depicts a mass burial site in Kahramanmaraş, Turkey” after the 7.8 magnitude earthquake that struck in 2023. “This specific aerial perspective became one of the most widely shared images of the disaster,” Gemini says, “illustrating the sheer scale of the loss.”
Seeing the identical burial picture on social media, others turned to X’s AI assistant Grok to test its veracity. Like Gemini, Grok will breezily guarantee you the photo shouldn’t be from Iran in any respect – though it lands on a unique date, catastrophe and placement. The picture is “from Rorotan Cemetery in Jakarta, Indonesia – a July 2021 stock photo of Covid mass burials. Not Minab,” it says.
In each circumstances, the AI solutions sound positive: they don’t equivocate, and even present “sources” for the unique picture, do you have to select to test them. Follow the thread to look at these, nonetheless, and also you’ll start to hit lifeless ends: both the picture doesn’t seem in any respect, or the hyperlink offered is to a information report that doesn’t exist. For all their impression of readability and precision, the AIs are merely mistaken.
The cemetery picture, it seems, is genuine. Researchers have cross referenced the photo of the web site with satellite tv for pc photographs that affirm its location, and it may be cross-referenced once more with dozens extra photographs taken of the identical web site from barely totally different angles, and once more with video footage – none of which specialists say present indicators of tampering or digital manipulation. The “factchecks” by Gemini and Grok are only one instance of a tidal wave of AI-generated slop – hallucinated info, nonsense evaluation and faked photographs – which might be engulfing protection of the Iran warfare. Experts say it is losing investigative time and dangers atrocities being denied – in addition to heralding alarming weaknesses as folks more and more depend on AI summaries for information and knowledge.
From the opening days of the warfare, factcheckers have been saved busy with a continuing circulate of faked imagery on-line. A photo of what the Tehran Times claimed was satellite tv for pc imagery of a US radar destroyed in Qatar was exposed as an AI fake made out of outdated Google Earth photos – its giveaways included the vehicles, which had been all in similar positions to the picture from two years earlier.
Widely circulated photographs of Khamenei’s physique being pulled from rubble had “tells” together with duplicate limbs amongst the rescuers. “One fake that stood out to me claimed to show a senior Iranian commander walking around Tehran disguising himself as a woman to avoid potential assassination,” says Shayan Sardarizadeh, a senior journalist at the BBC Verify staff, which makes use of forensic strategies to verify data and conduct visible investigations. “The street, the building in the background, and the surroundings all seemed like a realistic scene in Tehran.”
Sardarizadeh says AI now makes up a big portion of all of the misinformation the staff debunks – and the quantity is growing. In the first few weeks of the Gaza or Ukraine wars, for instance, most pretend posts the staff noticed had been outdated or unrelated movies, or repurposed online game footage. Now, “nearly half, if not more, of all the viral falsehoods that we now track and debunk are generative AI”.
That has partly been pushed by the ease with which anybody can now generate a practical video or photo. But the different huge shift is in folks utilizing AI to summarise the information or reply questions, moderately than going on to the unique supply. Google AI summaries and Grok had been solely rolled out to the wider worldwide public in mid 2024, and have quickly change into widespread: 65% of folks report regularly seeing AI summaries of information or different data, and the portion of individuals who say they’re utilizing generative AI to get data doubled in the past year. Often, nonetheless, AI summaries are merely mistaken. An international study in 2025 discovered about half of all AI-generated summaries had at the least one important sourcing or accuracy concern – with some instruments, comparable to Google’s common Gemini interface, that rose to 76%.
In the case of the Iran warfare, factcheckers say they’re seeing a deluge of this type of deceptive materials. As effectively as the Minab graveyard photographs, examples embody Grok inaccurately suggesting to X customers that video footage of fires in Tehran was actually from LA in 2017, and customers citing “AI analysis” to misidentify a missile filmed falling subsequent to the Minab college (quite a few munitions specialists say it is a US Tomahawk, a discovering strengthened by fragments reportedly discovered at the scene and inside US briefings on the bombing).
“Factcheckers now regularly have to address both a false post and also a misleading claim made by a chatbot in relation to that post,” Sardarizadeh says.
Part of the drawback is how LLM AI fashions (comparable to Grok, ChatGPT and Gemini) work. At a really primary stage, they’re probabilistic language fashions, developing sentences piece by piece primarily based on which subsequent phrase has the highest probability of being applicable. While that course of produces convincing, authoritative-sounding sentences, it doesn’t imply the AI has really analysed the materials in entrance of it.
“AI is perceived as an omniscient entity with access to everything, but without emotions,” says Tal Hagin, an open-source intelligence analyst and media literacy educator – so folks are likely to belief it. “What you are using is actually a very advanced probability machine, not a truth box.”
The drawback is compounded by the authoritative approach AI tends to current its findings. It will generate detailed “reports”, together with names and dates, references and sources – the variety of materials that means deep analysis and understanding, however could in reality be hallucinated or nonexistent. When the Guardian queried Gemini’s reply on the Minab {photograph}, saying “I don’t think that’s correct, can you search again?” it revised its discovering – however to a different incorrect location and 12 months. “I apologise for the oversight. Upon re-examining the image … this image was taken in Gaza in November 2023,” it says. Told that that reply was additionally incorrect, and the photo was from Iran, the bot revised once more – to Tehran, throughout the Covid pandemic. Told that the {photograph} was taken in Iran in 2025, it responded that it was from the aftermath of an earthquake in southern Iran.
X and Google didn’t reply to a request for remark. Both platforms’ AI companies observe of their small print that they could produce inaccurate outcomes.
For these investigating human rights abuses, the development poses new challenges. Chris Osieck, an impartial open-source investigator who has conducted investigations right into a quantity of civilian casualty bombings in Iran, stated researchers’ time was being wasted debunking AI materials. Debunking AI movies, for instance, usually entails rigorously inspecting them body by body for visible discrepancies. “That time should be devoted to what matters most: reporting on the impact this brutal war has on the people caught in the crossfire.”
And in circumstances comparable to Minab, the place the materials is demonstrably actual? Researchers worry the wash of AI slop is sowing doubt in folks’s minds that the atrocity they’re seeing proof of ever occurred in any respect. “As the technology continues to get better, it could muddy the waters so much that videos and images of real atrocities get dismissed as fake or AI,” say Sardarizadeh.
“I’ve already seen examples of this in relation to the conflicts in Gaza and Ukraine,” he says.
For those that have misplaced family members, accountability dangers being overshadowed by a mass of misinformation, suspicion and doubt.
“At the end of the day, one should also consider what this looks like from the perspective of the families of those who were killed,” Osieck says. “Imagine losing a child and then seeing AI being used online to claim that the event did not happen. That is not just an obstacle for investigators. It is also deeply disrespectful to the loved ones who are grieving.”
