The Iran War Fake Video Flood: How Games, AI, and Old Footage Rewrote a Real Conflict

A 21-second clip captioned "An Iranian plane VS a US ship" reached 7.3 million views on X after being posted around March 1, 2026. The "warship" is the USS Tennessee, a US dreadnought built in the 1910s and decommissioned in 1947; the "Iranian fighter jet" is a Messerschmitt Me 163B-1a Komet — a Nazi German aircraft from World War II. It's gameplay from War Thunder, a military simulation video game. ^[5]

Gaijin Entertainment's head of PR, Konstantin Govorun, confirmed to press: "Yes, this looks like War Thunder footage." ^[5] Texas Governor Greg Abbott had reshared the clip captioned "Bye bye"; he deleted the post without acknowledging the error, and his office subsequently put out a statement backing the military action ("Texas stands with President Trump"). ^{[6] [31]}

A video captioned "THIS IS TEL AVIV. THANK YOU IRAN!" spread on X around March 1, 2026. The footage is from the August 2015 chemical-warehouse explosion in Tianjin, China — one of the deadliest industrial disasters in recent Chinese history, which killed 173 people, including 104 firefighters, and injured hundreds more. This is at least the second time this specific clip has been weaponized in an unrelated war: it was already circulated, within days, as footage of Russia's February 2022 invasion of Ukraine. ^{[1] [2] [30]}

A separate, genuine video — real footage from the twelve-day Israel-Iran war of June 2025 — was also mislabeled as new 2026 combat footage, a reminder that confusion during this war ran in both directions: real footage mistaken for new, and old fakes mistaken for real. ^[33]

Football fans launching fireworks after Algeria's CR Belouizdad won the national championship in August 2020 was recirculated as Iranian missiles striking central Tel Aviv. The clip was traced to a Berbere Television post from August 7, 2020, geolocated to Al Mokrani Square in Algiers — and had already done a round of duty in 2023 as supposed footage of an Israeli strike on Gaza. ^[13]

Multiple X posts showing what appeared to be a US aircraft carrier sinking, attributed to Iranian missiles, accumulated more than 20 million combined views. The footage is from Arma 3, a photorealistic military simulation game, and its deck markings don't match any US carrier. ^[7] A second Arma 3 clip, captioned in Chinese as showing "the largest airstrike in modern history" by US F-15s, drew 5 million-plus additional views. ^[9] US Central Command separately rejected a related claim, pushed by Iran's semi-official Mehr News agency, that four Iranian ballistic missiles hit the USS Abraham Lincoln — saying the ship "was not hit" and the missiles "did not even come close." ^[8]

An October 1, 2015 fire at an apartment building in Sharjah, UAE was shared as evidence of Iran striking a supposed CIA facility in Dubai; there is no CIA headquarters in the depicted building. ^[10] Sharjah's civil-defense incident commander at the time said 19 people were treated for smoke inhalation and 250 households were affected by the original blaze. ^[11] Pro-Assad commentator Maram Susli ("Syrian Girl") was among those who amplified the clip as an Iranian strike on the "CIA headquarters." ^[12]

Tel Aviv ballistic missile strike (security cam): AI-generated video with duplicated building rooftops and unnatural orange smoke coloring, spread across X, TikTok, Instagram and YouTube. ^[9]

"Bahrain USS Navy Fleet Destroyed": an AI-generated clip showing a burning high-rise, with gibberish, Arabic-free signage and unnaturally uniform smoke rather than the diffuse plume of a real strike. It was first posted by an account that disclosed the clip as AI-generated for "entertainment purposes," then reshared without that disclosure and amplified by Iran-aligned accounts. ^{[16] [17]}

Captured US soldiers in Tehran: AI-generated images claiming to show US personnel captured by Iran retained a visible Google Gemini watermark — a diamond mark in the corner — that went unnoticed by viral sharers; Google's SynthID detector confirmed synthetic origin. No credible reporting corroborates any such capture. ^[15]

Khamenei body under rubble: AI-generated images claiming to show Khamenei's body under rubble appeared within hours of his confirmed death in the February 28 strikes (Iranian state media confirmed the death March 1). Full Fact and AFP identified the images as synthetic via a SynthID watermark. ^{[12] [14]}

Tehran Times satellite imagery: a "before vs. after" image pair claiming to show a destroyed US radar at Al-Udeid Air Base in Qatar. Analyst Tal Hagin identified the "before" image as a Google Earth image from February 2, 2025 — a year before the war — noting that parked cars sat in identical positions in both frames; AFP's SynthID scan also turned up gibberish coordinates embedded in the fake. ^{[3] [8]}

The Associated Press documented that Iran-conflict disinformation wasn't merely opportunistic: it followed at least two organized, state-linked tracks operating during the war itself, with a third — an Israel-aligned AI campaign — standing as a documented precedent from the previous year's fighting rather than a confirmed feature of this one. ^[16]

Iran's semi-official Tasnim news agency, citing IRGC statements, reported that 650 US troops were killed or wounded in the war's first two days; CENTCOM put the real figure at six killed. Mehr News reported that four Iranian ballistic missiles struck the USS Abraham Lincoln; CENTCOM said the missiles didn't come close. Tehran Times published the fake Al-Udeid radar imagery within hours of the initial strikes — an operational tempo suggesting pre-prepared material. ^[8] Separately, Iran-aligned accounts amplified the AI-generated Bahrain "fleet destroyed" clip after it began circulating without its original AI disclosure. ^{[16] [17]}

A year earlier, during the twelve-day Israel-Iran war of June 2025, the University of Toronto's Citizen Lab had already documented an Israeli-linked AI influence operation it designated "PRISONBREAK," using more than 50 inauthentic X profiles to push regime-change messaging inside Iran — including an AI-generated video of an Evin Prison bombing posted within roughly an hour of the real airstrike. Citizen Lab assessed the operation as "most likely undertaken by an entity of the Israeli government or a private subcontractor working closely with it." ^[19] Whether the same operators were active during the 2026 war is not established by any source checked for this piece, and the claim is included here only as documented historical context, not as a confirmed feature of Operation Epic Fury.

Russia's "Operation Overload" (also tracked elsewhere as Matryoshka or Storm-1679) took a two-layer approach to the 2026 war. One layer was direct: a warning falsely attributed to Israeli intelligence, telling Israelis in Germany and the US to avoid public places. ^[16] The other was parasitic: the same network fabricated entire cloned Euronews broadcast segments, using AI voice cloning of real Euronews journalists, to spread claims that "Ukrainian looters" attacked Dubai shops after the strikes, that authorities detained 19 Ukrainians over it, and that Armenian PM Nikol Pashinyan owned $170 million in UAE property — distributed through the pro-Kremlin Pravda network. Armenia's press secretary called it "a classic FIMI [Foreign Information Manipulation and Interference] mechanism." ^[26]

AI-generated content was one of the largest categories of documented fakes in the Iran war, alongside recycled archival footage and video-game clips. The tools involved — including Google Veo and Midjourney — are commercially available and require no technical expertise; PolitiFact noted that Veo-generated clips are partly recognizable by their roughly eight-second length limit. ^[3]

Open-source intelligence analyst Tal Hagin told PolitiFact: "These fabrications are becoming more convincing and harder for seasoned experts to identify." ^[3] Rolling Stone likewise reported disinformation researchers' alarm at the volume of AI content overwhelming crisis reporting during the war. ^[23]

The most-documented failure of AI amplifying disinformation was Grok, X's built-in chatbot. Asked about the AI-generated Tel Aviv ballistic-missile video, Grok told users: "No, this isn't AI, it's a real photo from today's Iranian ballistic missile strikes on central Israel" — falsely citing Reuters, CNN and Euronews as sources. ^[9] Separately, Full Fact caught Grok insisting that genuine footage of a Glasgow, Scotland fire showed "firefighters tackling a major blaze in a Tel Aviv building." ^[28] RTÉ Primetime documented this as a pattern across at least four distinct query types during the conflict, including conflating a real Iran strike aftermath with an unrelated 2021 Kabul attack and misreading Tehran imagery as Italian. ^[27]

On March 5, after at least some of these failures had already been publicly documented, Elon Musk posted: "Use Grok to fact check and ask questions about any post." He did not retract this endorsement, and X did not respond to RTÉ's inquiries about Grok's reliability during the conflict. ^[27]

AI images with a visible Google Gemini watermark circulated as "captured US soldiers in Tehran" — the AI logo present in the frame itself, apparently unnoticed by viral sharers. ^[15] Tehran Times' fake satellite imagery was exposed using Google's SynthID detection tool, which turned up gibberish coordinates embedded in the image alongside the naked-eye tell that parked cars sat in identical positions in the "before" and "after" frames. ^{[3] [8]}

X's head of product, Nikita Bier, said that "in 99% of cases, it's just people looking to game monetization" when asked about AI-generated conflict videos on the platform. ^[29] X's creator program pays roughly $8–12 per million verified-user impressions to accounts with an active Premium subscription and at least 5 million organic impressions in the prior three months — a widely reported general feature of the program, not specific to this war. Under that formula, a video reaching 20 million views, like the Arma 3 carrier clip, would fall in a rough $160–240 range, though X does not publish per-post payouts. ^[29]

Maldita.es had already documented, in a report published a month before the war began, an industrialized version of this business model on TikTok: 550 coordinated accounts that had posted more than 5,800 AI-generated videos about protests and unrest across 18 countries, drawing 89 million-plus combined views. The report wasn't about Iran specifically — it's evidence that the monetized-fabrication infrastructure already existed and was simply redirected once the war began. ^[24]

X announced 90-day suspensions from creator revenue-sharing for undisclosed AI-generated conflict videos on March 3, 2026, with permanent bans threatened for repeat offenders. The policy relies on creator self-disclosure, and researchers monitoring the platform afterward said their feeds "were still flooded with AI-generated content about the war." ^[29]

Google's SynthID AI-detection tool was used by fact-checkers including AFP and Full Fact to identify AI-generated images during the conflict, including the Tehran Times satellite imagery and the Khamenei rubble images. It only detects content generated by Google's own AI tools — images from Midjourney, Stable Diffusion, RunwayML or other non-Google platforms fall outside its coverage, a structural blind spot in current detection infrastructure. ^{[3] [12] [14]}

A separate, previously undocumented category: hundreds of synthetic videos depicting women as Iranian soldiers — a fabrication at odds with Iran's actual policy barring women from combat roles. Combined views reached 10 million-plus across platforms; TikTok banned 52 accounts distributing the content. ^[25]

The 2026 Iran war is at least the third major conflict in which War Thunder and Arma 3 footage has been widely shared as authentic combat footage, following the 2022 invasion of Ukraine and the 2023 Israel-Hamas war. Gamereactor documented War Thunder, Arma 3, and Call of Duty footage all circulating as Iran-war footage in the same news cycle — including a White House-posted video that opened with Call of Duty: Modern Warfare III gameplay before cutting to real strike footage. ^[21]

Bohemia Interactive, the Czech studio behind Arma 3, used nearly identical language after this war to what it said following the 2023 Israel-Hamas war: "It's disheartening for us to see the game we all love being used in this way." ^[22] Its documented approach has been cooperation with fact-checkers like AFP and Reuters rather than technical fixes such as watermarking or output restrictions.

The War Thunder incident is particularly instructive. Abbott's non-response — deletion without acknowledgment, followed by a statement backing the military action — illustrates the political utility of fake war footage even after exposure: it generated engagement and reinforced a narrative regardless of its authenticity. ^{[6] [31]}

The Iran war disinformation surge was not simply "more of the same." Three conditions in convergence made it qualitatively different from prior documented conflicts.

Iran's 98% internet traffic collapse, recorded by Cloudflare on February 28, is historically rare. In Ukraine, citizen journalists provided continuous ground-truth footage from Day 1. In Iran, the near-complete blackout meant fabrications faced essentially no competition from verified reality. ^[18]

Human Rights Watch documented that the shutdown "help[s] conceal large-scale atrocities, contribute[s] to the spread of mis- and disinformation, and unlawfully restrict[s] access to information" and "severely hampers the work of journalists and human rights monitors, including documentation and reporting on possible laws of war violations by all parties." ^[18]

The tools used for fakes in this conflict — Google Veo, Midjourney, and others — are commercially available and require no technical expertise, producing outputs convincing enough that a sitting US state governor and a major platform's own AI chatbot could not immediately identify them as fake. ^{[3] [20]}

X's creator economy and TikTok's engagement monetization pay some creators more for viral fake conflict content than accurate reporting typically earns. Accuracy requires investment — research, verification, time; fabrication is comparatively instant. The pre-existing 550-account, 5,800-video TikTok operation Maldita documented represents a functioning disinformation business model that predates this specific war. ^[24]

The Tianjin 2015 explosion footage has now been recycled in at least two separate, unrelated conflicts over 11 years: the 2022 invasion of Ukraine and the 2026 Iran war. It's a data point illustrating the durability of viral fabrications once uploaded — the video was never fully purged from social media after its 2022 debunking, making its 2026 resurgence possible. ^{[1] [2] [30]}

Across Poynter and PolitiFact, a consistent toolkit emerged for identifying fake conflict footage: reverse-image search to check whether content predates the war; watching for AI artifacts like distorted text, impossible textures, or duplicated background elements; checking Google Maps against depicted street signs and geography; comparing structural details (e.g., a carrier's actual deck layout) against the real thing; running suspect images through Google's SynthID checker; and, for suspected AI video, noting that some generators cap clips at roughly eight seconds. When in doubt, established fact-checkers — not social platforms' own AI chatbots — remain the more reliable check. ^{[1] [3]}

The lesson from War Thunder, Tianjin, and an Algerian football celebration is not simply that people are credulous — it's that platform incentive structures reward fabrication and punish verification. Fabricating a fake carrier-sinking video takes minutes and can plausibly generate roughly $160–240 in X creator revenue before any label appears. Verifying authentic footage from inside Iran is close to impossible when the country's internet is 98% dark. ^{[18] [29]}

At least two state-linked actors operated disinformation campaigns during the war itself — Iran manufacturing fictional victories for domestic consumption and foreign amplification, and Russia parasitically recycling the conflict to serve an unrelated anti-Ukraine narrative — with a third, an Israel-aligned AI campaign, documented as a precedent from the prior year's fighting rather than confirmed for this war specifically. All of the AI tools involved were commercially accessible to anyone. ^{[16] [19] [26]}

The Tianjin explosion footage — now debunked across two separate, unrelated conflicts spanning 11 years — is the period at the end of this sentence. Fake war footage rarely fully disappears; it waits for the next war. The 2026 Iran conflict was that war for more than a dozen pieces of fabricated content simultaneously, amplified by AI tools that generate new fakes faster than fact-checkers can debunk the old ones, on platforms whose financial logic makes the problem not merely persistent but profitable. ^{[1] [20]}