For a long time, media rested on a simple principle: A news story backed by images or videos seemed more credible. A picture from a crisis zone, a video of a police encounter or a political leader speaking gave a news report an added layer of credibility by allowing the consumer to see the “evidence” for themselves. However, the advent of artificial intelligence and AI-generated and manipulated content is changing the scenario. The media can no longer simply present an image or video as evidence.
The 2020 video of George Floyd’s murder became a powerful example of how video footage can help establish credibility. The recording clearly showed former Minneapolis police Derek Chauvin kneeling on Floyd’s neck as he repeatedly said he could not breathe. As Floyd’s aunt Angela Harrison says, “The video shook the world. Everyone saw the brutality unfold. There was clear evidence, and it became a catalyst for the Black Lives Matter movement.”
When video footage does not necessarily tell the whole story
Six years later in January 2026, another such encounter produced a video, that of Renee Good who was fatally shot by an Immigration and Custom Enforcement (ICE) officer in during a federal immigration enforcement operation in Minneapolis. But something different happened here. Unlike the Floyd footage, the Good video did not provide a clear narrative of what had happened before or after the ICE agent shot her. There were multiple accounts and divergent public arguments, but the video did not settle the key question of what had happened in the moments before the shooting. The Good case in still under investigation.
“These ICE videos do present irrefutable facts — a woman drove her car and then she was shot dead by an ICE agent,” said Duy Linh Tu, a documentarian and professor at the Columbia University journalism school. “What the videos can’t show is the intent of the woman or the officer. And that’s the tricky part.”
“The video of the incident has been generally inconclusive about whether Good’s vehicle actually hit Ross before he opened fire. “Even if she did, many experts question whether that represented grounds for firing his weapon. Clearly, however, that would bolster public sympathy for the officer,” the AP reported.
Frankie McLister, a reporter at WCCO-TV, Minneapolis who covered Good’s death and Operation Metro Surge says: “George Floyd’s death put police departments under a microscope around the world, while Renee Good’s death became a catalyst for an uprising against a surge of federal immigration enforcement in Minnesota. In both cases, videos of the incidents circulated online almost immediately, becoming central to how the public understood what had happened.”
AI can complicate the relationship between images and credibility
The larger question however is what happens to this link between images and credibility in the age of artificial intelligence (AI). Images and videos often conveyed trustworthiness though they were never necessarily completely free from manipulation. Realistically, they could be photoshopped, edited, taken out of context or presented in a manner that shaped the understanding of an occurrence. However, there was still an assumption that what the footage showed had happened. AI has complicated that assumption.
For example, in the run up to 2024 presidential elections, Donald Trump posted on Truth Social an AI generated image of Taylor Swift in an Uncle Sam outfit, falsely claiming she had endorsed him. The image was clearly fake and was accompanied by other depictions, some also apparently AI-generated, of young women in T-shirts reading “Swifties for Trump.”
Similarly, an October 2024 audio clip of Trump issuing a “rally cry” to Britons attending a far-right protest in London was found to be AI-generated, according to a forensic analysis tool and three experts who spoke to Reuters.
The new challenge of verifying videos and images
Some media experts have argued that in this age of AI-generated or manipulated content, the credibility of videos such as the one involving Good can be questioned, and that such footage has inherent limitations. In these AI times, journalists have a tougher task: they must verify the credibility of their sources and the authenticity of the images and videos they use to support the narrative.
However, that task may be harder than it seems, at least for now.
The Tow Center tested seven AI platforms for their ability to identify the provenance of photographs and found all seven models failed to consistently correctly identify where the photo came from.
