A wall of orange rises behind a row of dark towers. Someone filming from a high-rise swears, then runs. The clip looks exactly like the kind of phone video that reaches social media before a correspondent can verify it.
In July 2026, it arrived with a fresh location. Text laid over the video called it a “massive violent explosion” in Bushehr, Iran, following a US airstrike. Full Fact reported that one Facebook version had received more than 1,500 likes.
Verdict: False. Confidence: high. The footage was recorded in Tianjin, China, on 12 August 2015. It shows a fire and explosions at a hazardous-goods warehouse. It does not show Bushehr, a US projectile or Iran in 2026.
The video is real. The claim attached to it is false.
Eleven years older than the caption
The quickest way to break this claim is to find an older copy.
BBC News published the same eyewitness sequence on 14 August 2015, identifying Dan Van Duren as the person filming. ABC News reported at the time that Daniel Van Duren had uploaded the video on 12 August, the night the explosions hit Tianjin.
That contemporary record predates the Bushehr caption by almost eleven years. It also supplies details that fit the footage: a port skyline, a hazardous-goods site and an English-speaking expatriate filming from a nearby high-rise.
China's final State Council investigation record identifies the site as Ruihai International Logistics' dangerous-goods warehouse in Tianjin's Binhai New Area. The published toll was 165 people killed, eight missing and 798 injured.
Those numbers restore something the viral caption removes. This was not anonymous spectacle waiting for a new geopolitical label. It was a specific industrial disaster with victims, investigators and a date.
One fireball, several wars
The Bushehr post is the newest documented costume, not the first.
In February 2022, the footage was shared as a fire caused by a Russian airstrike in Ukraine. The person recording speaks English with an American accent, a clue that helped researchers trace the older source.
In March 2026, the same scene became explosions in Tel Aviv, sometimes attributed to an Iranian missile attack.
In June, it returned to Ukraine with a bigger invention: a Russian attack that had supposedly killed President Volodymyr Zelenskyy. The footage was still from Tianjin. Zelenskyy was alive and appearing in public.
Then came Bushehr. That caption borrowed a real event from the news cycle. US projectiles did strike points in the city on 14 July, according to reporting based on local and state media. A real strike made an unrelated explosion look plausible.
The sequence is now a travelling visual asset: Tianjin 2015, Ukraine 2022, Tel Aviv 2026, Ukraine again, then Bushehr. Each caption asks the same old disaster to testify about a new war.
Reality is useful raw material for a lie
Synthetic media deserves scrutiny. This example shows why verification cannot begin and end with the question “Was this made by AI?”
No generated fireball is needed when the internet already contains a frightening one. Authentic footage brings natural motion, imperfect camera work, human panic and the physical complexity that fake-video detectors look for. The source is genuine enough to survive a viewer's first suspicion.
The falsehood lives in the metadata people can see: the caption, the date and the location. A platform strips away much of the original file history, a user adds a few words, and a Chinese industrial disaster becomes evidence for whichever conflict dominates the feed.
This is why “the video is real” is an incomplete verdict. A video claim has at least two parts: what the pixels show and what the publisher says those pixels mean. Either part can fail.
The distinction also protects against the opposite error. Calling the clip “fake” can imply that the explosion was fabricated. It was not. People in Tianjin lived through it. The accurate correction is narrower and stronger: the footage is authentic, and the Bushehr identification is false.
How to verify a recycled video
You do not need to recognize Tianjin on sight. You need to make the clip compete with its own past.
1. Freeze the most distinctive moments. Use an early view of the skyline, the first fireball and the frame before the filmer moves. Tools such as InVID can extract keyframes, but ordinary screenshots also work. 2. Search the frames, not the caption. Run several reverse-image searches. The caption supplies the claim under examination; repeating its keywords often returns copies of the same rumour. 3. Move backwards in time. An upload from 2015 defeats a claim that the scene was filmed in 2026. Look for the earliest dated appearance, not the version with the most views. 4. Corroborate the old identification. A matching frame is a lead. Contemporary reporting, the filmer's identity and an official incident record establish the event. 5. Check the real-news kernel separately. A strike may genuinely have happened in the named city. That does not authenticate every dramatic video posted beside it.
AFP described using reverse searches on frames to trace another Ukraine version to 2015 coverage by the BBC, the Guardian and ABC. Full Fact's verification guide walks through the same sequence and the American-accent clue.
This method still works after a video has been cropped, subtitled or recompressed. You are identifying the underlying scene, not proving that every circulating file has identical bytes.
The caption keeps the power and discards the dead
We could not independently retrieve the current Facebook reel from this environment. Full Fact preserves its wording, date and reported engagement. That is enough to test the public claim against dated evidence, but not enough to judge what the original poster knew or intended.
Intent is not required for the verdict. A person can repeat a false caption sincerely. The result still moves a real disaster across borders and years, teaching viewers that any spectacular image can be recruited as eyewitness evidence.
Tianjin's explosion keeps returning because it looks undeniable. The right response is not to deny the image. It is to restore what the caption stole: the place, the date and the people whose catastrophe it was.
The fireball is real. The new war is the lie.