An ICE agent shot Renee Nicole Good dead on a Minneapolis street, and bystanders filmed the scene from several angles. The clips did not agree. One showed the agent standing beside a vehicle whose wheels pointed away from him. Another, filmed from a balcony and later shared by President Trump, made it look as though the vehicle had struck him. A newsroom cannot settle that kind of dispute by watching a video once. It settles it by pulling the video apart.
more
How a newsroom pins down where a clip was filmed
The starting technique is geolocation, which pins down where a clip was filmed. Visual investigators pair it with satellite imagery, archived social media posts, and other user-generated content to build a timeline around the clip. Independent groups pioneered these methods. They include Bellingcat, Storyful, SITU Research, Forensic Architecture, Airwars, and Syrian Archive. The methods gained prominence during the Arab Spring in 2011 and the Syrian civil war, when reporters on the ground faced serious danger. Major outlets later built their own teams. Among them are The New York Times, The Washington Post, The Wall Street Journal, Le Monde, CNN, the BBC, and the Financial Times. The work has won Pulitzers, Peabodys, and Emmys.
Geolocation answers where. A second technique answers when, and what order things happened in. In the Minneapolis case, The New York Times synchronised the balcony clip with a longer, clearer street-level video filmed by bystander Caitlin Callenson. That let it check the sequence of events frame by frame. Its analysis found that when the agent fired, he stood to the left of the vehicle with its wheels turned away from him. That conflicted with claims that the vehicle had rammed or nearly struck him. The Washington Post ran its own frame analysis under a “Visual forensics” byline, and Bellingcat examined the footage to track the position of the agent’s gun and phone in the seconds after the shooting.
Why the same video read different ways
Newsrooms without that capacity fell back on describing the clips in prose, and their descriptions diverged even though the underlying footage was identical. Fox News wrote that Good is heard telling agents to “go around.” Outlets also had to choose whether to quote the agent as saying “Get out of the car” or a cruder version of the line. Ståle Grut, a researcher at the University of Oslo, had predicted the gap a year earlier. Newsrooms that had not built visual investigative competence, he said, would pay for it “in lengthy corrections, lost credibility, and missed opportunities.”
When the technique itself produced the wrong answer
Verification is not infallible, and its record includes a public failure. After an explosion struck Gaza City’s Al-Ahli Arab Hospital complex on October 17, 2023, several visual-investigation teams analysed the same bystander footage and reached conclusions that contradicted one another. Corrections, editors’ notes, and public admissions of error followed at multiple outlets. The episode showed that a technique built to resolve uncertainty can, under the pressure of a fast-moving story, produce more of it.
The new problem: proving a real image is real
Artificial intelligence has added a harder problem than confirming a fake. In April, Planet Labs stopped releasing daily satellite imagery of the Middle East, narrowing the record investigators could check against. In July, Google added image-generation tools to Google Earth that let anyone produce a convincing fake satellite image, and only reversed the feature after journalists objected. Imogen Piper, a visual-forensics reporter at The Washington Post, put the shift this way. She worries less about proving fakery to a reader now, she said, and more about how “we prove to people that something is real.”
Alexa Koenig, who directs the Investigations Lab at UC Berkeley’s Human Rights Center, described the same problem from the reader’s side. The signals that once told an audience a piece of footage could be trusted no longer carry that weight, and she said the field has “got to figure out how we pivot in this moment.”
That leaves newsrooms leaning on what worked in Minneapolis: synchronise every version in circulation, and say plainly where the evidence runs out.
