The protests in Iran are actual. The nation’s financial desperation runs deep, and hundreds of thousands of residents wish to see a corrupt and repressive regime gone. The violent crackdown on the protests can also be actual and seems to have price 1000’s of lives. But the accounts, images, and movies popping out of Iran are riddled with accusations of AI manipulation and fakery which have the impact of calling even what’s true into doubt.
There’s a time period of artwork for the profit that accrues to unhealthy actors after they sow public uncertainty as as to whether something in any respect is actual: It’s the liar’s dividend, and it’s very a lot in proof in Iran.
AI-generated or -enhanced on-line content material has turn into a worldwide drawback for these searching for to know or doc protest actions, as a result of it permits events to form narratives. In 2025, AI-generated photographs of actual protest moments muddied the view of occasions in Turkey, and AI-generated content material grew to become a mobilization instrument for Nepal’s Gen Z rebellion. However Iran is maybe essentially the most fraught enviornment of all. A number of political factions, international governments, and the Iranian regime itself are all competing to form the narrative of present occasions. And the instruments for fabricating actuality have by no means been as superior and accessible as they’re right this moment.
Movies of Iran circulating on social media present crowds moved by an array of convictions and needs. Some merely chant in opposition to the regime; others chant for Reza Pahlavi, the crown prince of Iran’s deposed monarchy, or invoke his father, the final shah. Quite a few confirmed movies present demonstrators venting financial anguish and hatred for Iran’s supreme chief, Ayatollah Ali Khamenei; the safety forces; and the Islamic Republic itself. All of those views of the protests are grounded in Iran’s advanced actuality. But the presence amongst them of photographs and movies which were—or are even suspected of being—doctored with AI, mislabeled, misattributed, or spliced with content material from different locations or occasions has given oxygen to the darkish imaginings of the motion’s enemies inside the regime.
As a diaspora Iranian who has studied on-line deception, censorship, and surveillance efforts for nearly 15 years, I take a selected curiosity in Iran’s data surroundings. On the human-rights group Witness, I’m concerned with one thing known as the Deepfake Fast Response Power, which helps journalists and human-rights defenders assess manipulated media. We’ve a lot to work with through the present Iran disaster.
What we’re seeing shouldn’t be a single disinformation marketing campaign however one thing a lot messier. Iran’s data ecosystem has been formed by excessive mistrust of the Islamic Republic’s official media; social-media affect efforts coming from each the regime and international powers; and rampant distortions of on-line content material—some misleading, some careless, some properly intentioned—that make verifying what Iranians are literally saying or doing throughout these protests tough. The liar’s dividend redounds to the regime greater than anybody else.
Iran’s nationwide rebellion erupted on December 28. And inside hours, regime-aligned accounts started dismissing genuine photographs from the protests as AI-generated.
An early, high-profile case concerned an image that got here to be referred to as Iran’s “tank man” photograph. Its supply was a low-quality video that had been captured with a zoom lens on December 29, displaying a protester going through down safety forces in Tehran. A BBC Persian journalist posted a screengrab from the video the identical day, and that picture, which had the next image high quality, started to flow into. The occasion was actual: It has been verified from a number of angles and by fact-checkers. However someplace alongside the road, somebody enhanced the nonetheless picture of it with AI enhancing instruments, more likely to make the blurry authentic screengrab sharper and extra shareable.
Regime accounts seized on seen artifacts from AI enhancement as alternatives to dismiss the photograph and different footage from the protests. “It’s all AI slop,” one regime account with a visibly AI-generated profile photograph opined. “Simply if you assume Zionists can’t be extra pathetic, they show you incorrect.”
Deepfakes have given AI a fame for being misleading. However many generally employed image-editing instruments use generative-AI capabilities, for higher or worse. The general public doesn’t usually know methods to distinguish good-faith use of those instruments from malicious fakery. And this enables bad-faith actors such because the Islamic Republic, which has spent many years undermining protests and dissidence, to make use of not solely AI itself but in addition the suspicion it generates as an accelerant.
The identical sample has performed out with audio manipulations. On December 30, an account presenting itself as affiliated with the Mujahedin-e Khalq—a controversial exiled opposition group referred to as the MEK that has an armed wing and was as soon as designated a terrorist group by america—posted a video during which pro-Pahlavi chants are clearly dubbed over protest footage. The account then “uncovered” the video as manipulated and used it to assert that monarchists have been fabricating help. Inside hours, pro-regime accounts had reposted this content material to amplify the accusation that pro-Pahlavi chants have been pretend.
What was putting was that the alleged MEK account each launched the manipulated video and was the primary to name it out. Furthermore, regime-aligned accounts have been able to rebroadcast that submit nearly instantly. Iranian researchers have lengthy documented the regime’s use of opposition personas on social media to sow confusion, divide the opposition, and discredit genuine documentation, and leaked paperwork and movies from inside the regime have confirmed the observe.
Inside a day of the supposed MEK submit being printed, a Russian propaganda account set the AI-enhanced “tank man” photograph alongside the allegedly manipulated Pahlavi audio and offered each as proof that the protests have been a international fabrication. The loop was closed: manipulated content material launched by what might have been a regime account, amplified by pro-regime voices, then aggregated by Russian state-aligned media as proof of an AI-driven Western psyop.
If that cycle isn’t sophisticated sufficient, contemplate that the regime’s accusations capitalize on at the very least one documented occasion of manipulation by monarchists. On December 29, Manoto, the London-based Persian-language channel aligned with the monarchist opposition, aired footage of crowds chanting in favor of Pahlavi. The audio is actual—however traceable to a memorial for Khosrow Alikhordi, a human-rights lawyer whose suspicious loss of life final month drew 1000’s of mourners. It had been lifted and overlaid onto unrelated protest footage. Whether or not the doctored video was planted by hostile actors or deliberately created by somebody inside the royalist camp, the impact was the identical: It handed malicious accounts precisely the precedent they wanted.
Royalists are usually not the one anti-regime camp that has both generated or circulated pretend content material that serves their narrative. On December 30, Radio Farda and different accounts shared a 2020 video, as if it have been footage of the present protests, of protesters chanting, “No to the oppressor, be it the shah or the chief.” BBC Confirm’s Shayan Sardarizadeh flagged the misattribution.
The Islamic Republic has a behavior of attributing genuine, homegrown dissent to Israeli and American conspiracies. This supposed international subversion then turns into a pretext for repression. The most recent disaster isn’t any completely different in that regard—however now international actors are serving to enhance the regime’s narrative by sharing doctored, deceptive content material that the regime can then expose.
Israel has been utilizing AI-generated content material to push anti-regime and pro-Israeli narratives in Iran since at the very least final summer time’s 12-day conflict. The Citizen Lab, a analysis and coverage heart primarily based on the College of Toronto, factors to AI-fabricated footage of a supposedly surgical strike on Evin Jail, amongst different examples. A Haaretz investigation discovered that Israeli affect operations have been selling concepts aligned with Iranian monarchist factions.
Shortly earlier than the protests broke out final month, an Israeli diplomat shared a montage of 4 alleged strikes on Iranian command facilities through the June conflict captioned, “very exact.” Our evaluation discovered that of the 4 clips, one is genuine footage that had aired on Iranian state tv; the opposite three present indicators of getting been generated by AI. BBC Confirm traced the video montage to an Instagram account that has greater than 120,000 followers and identifies itself as “Islamists and Khomeinists’ Nightmare” and “Media Warfare Knowledgeable.” When customers identified the manipulation of the air-strike footage, the diplomat didn’t delete the submit. He acknowledged that components of it could have been AI-manipulated, however argued that the video nonetheless demonstrates the precision of Israel’s operations—in different phrases, he stood by admittedly fabricated content material as a result of it illustrates a narrative he wished to inform.
Maybe equally, on January 1, the Israeli International Ministry posted a picture on its Persian-language X account of Iranian police blasting two protesters with a water hose. The picture was AI-altered: The authentic, printed by BBC Persian’s Farzad Seifikaran the day earlier than, reveals the protesters being hosed down from a distance after which an officer approaching with a loudspeaker, not a hose.
These types of interventions could also be meant to spice up the opposition to the Islamic Republic, however using doctored media has different, little doubt unintended, results. Actors with a vested curiosity in hiding what is going on in Iran’s streets use the existence of untrustworthy content material promoted by international powers to persuade the general public that all documentation of the protests is a international deception. The dividend once more accrues to the Islamic Republic.
Typically the issue shouldn’t be manipulation in any respect however simply the impossibility of capturing a fractured actuality. On December 29, Iran Worldwide shared footage from a march within the city of Malard, the place a crowd was chanting for Pahlavi. One other consumer posted a distinct angle of the identical road and the identical march, solely with completely different audible slogans. One consumer who had shared the primary footage issued a correction; others accused the Pahlavi camp of spreading falsehoods. However who’s to say that on that lengthy road, completely different crowds at completely different moments didn’t merely chant various things?
That is epistemic fog. Even with out deception, folks working in good religion, attempting to confirm what they’ve seen, can’t know what’s consultant. Such information is elusive in any case: Iran is a rustic of greater than 90 million folks whose aspirations can hardly be generalized from any single video or protest chant.
Journalists, courts, human-rights investigators, and protest actions in all places have lengthy trusted shared requirements of proof. At this time we’re seeing what occurs when these requirements collapse. When any picture will be dismissed as presumably artificial, these in energy not have to suppress speech. They’ve solely to boost questions that undermine perception in a shared actuality.
Since Thursday night, the Iranian regime has thickened the fog to close impenetrability with a nationwide web shutdown and a blackout of communication with the surface world. From the place I stand, the knowledge image has fractured additional. Verified footage has slowed to a trickle; what comes out largely passes by Starlink terminals. Diaspora Iranians are frantic, unable to achieve relations. Dying-toll figures fluctuate wildly, as documentation organizations wrestle to entry knowledge: The vary now sits between 12,000 and 20,000 lifeless, however the full scale is unknown.
Our Deepfake Fast Response Power has been attempting to assist validate reputable human-rights documentation that’s generally falsely tagged as AI. An unlimited effort is afoot amongst human-rights organizations to precisely rely the lifeless. In the meantime, regime accounts have been disseminating footage from counterprotests in favor of the Islamic Republic—clips that some opposition accounts recommend are AI-generated, and that official information media vehemently insist are actual. Everyone seems to be primed to mistrust the whole lot, and the shutdown ensures that nothing—not the size of the protests, the severity of the crackdown, or the state of the regime—will be witnessed in actual time.
The Iranian regime has spent greater than 4 many years undermining the credibility of dissent. Now it has AI not simply as a instrument for creating fakes however as a rhetorical weapon. Each glitch, artifact, and even protecting blur turns into proof that nothing will be trusted. In a world the place coverage makers and platforms put money into AI detection and fact-checking, maybe abnormal folks wouldn’t be combating by such darkness, or unhealthy actors reaping such dividends.
What Iranians need and are combating for is a query solely they will reply. And their reply deserves to be heard—not formed by international governments, dismissed by regime propagandists, or drowned in fog that others have helped create.
