For most of legal history, video was considered close to infallible. Juries trusted it. Judges relied on it. Video evidence could be a “smoking gun,” incontrovertible proof, dispositive. A recording, unlike a human witness, could not misremember or be coached. That assumption is now crumbling.
Around six months ago, a California case sent shockwaves through the legal world after plaintiffs were accused of submitting an AI-generated deepfake video as testimony. The judge caught it — but barely. Experts noted that the judge spotted glitches like repeated eye blinks and looped video, and warned that it may have been luck, because the technology is already beyond those telltale flaws.
The legal system has been scrambling to respond to advances in deepfake technology. In August 2025, the Advisory Committee on Evidence Rules published proposed Federal Rule 707 for public comment, which would extend the Rule 702 reliability standard to machine-generated evidence. The comment window closed in February 2026. Critics caution that the rule applies only to evidence that the proponent acknowledges was created by AI, and not to evidence whose authenticity is disputed — meaning it does little to help courts detect undisclosed deepfakes.
What the law hasn’t fully grappled with, however, is the psychological dimension of this problem. Because the threat isn’t just that fake evidence exists. It’s that our minds are extraordinarily bad at detecting it.
Why We Are Wired to Believe What We See
The famous line “Who you going to believe, me or your own eyes?” is a classic comedy quote popularized by Chico Marx in the 1933 Marx Brothers film Duck Soup. The humor lies in the assumption that humans should always believe their own eyes. Indeed, humans are visual creatures. Cognitive science has long established that we process visual information faster and with greater emotional weight than text or abstract argument. When we see someone on a screen say something, we experience a powerful sense of presence and authenticity — what researchers call the “veridicality heuristic,” the intuitive assumption that perception equals truth.
This is precisely why video evidence has always been so persuasive in courtrooms. Jurors are often influenced by the confidence and detail with which a witness recounts an event, equating a vivid, personal narrative with reliability and truth. Now imagine that same psychological pull applied to a fabricated video — one a person never made, showing them confessing, threatening, or committing something they never did.
Humans have proven to be poor judges of whether a digital artifact is real or fake, and technologies designed to detect AI-generated content have proven to be unreliable and biased. As discussed at length in a recent University of Chicago Legal Forum article, there is no foolproof way today to classify text, audio, video, or images as authentic or AI-generated.
The Deepfake Defense — and Its Mirror
Defense attorneys have begun invoking what’s being called “the deepfake defense,” a term that describes how the ease of producing deepfakes enables bad actors to dismiss true recordings as fabrications. This creates a deeply troubling psychological double bind in the courtroom: jurors may grow skeptical of genuine evidence while remaining susceptible to sophisticated fakes.
If jurors begin to assume that “anything can be AI,” legitimate recordings may lose their persuasive value. Legal scholars have noted that widespread skepticism about digital media could erode confidence in otherwise admissible evidence. The psychological consequence is a kind of epistemological paralysis — a courtroom where no one quite knows what to trust.
The stakes are concrete. As the National Center for State Courts noted in February 2026, a woman spent two days in jail after her ex-boyfriend allegedly fabricated AI-generated text messages that led to her arrest for violating a protective order. Prosecutors eventually dropped the charges, but only after eight months of legal proceedings.
The Problem of Contaminated Perception
Decades of memory research offer a useful parallel. Psychologist John Wixted of UC San Diego, whose work has influenced how police conduct lineups, has argued that eyewitness memory is no different than DNA or fingerprints in one crucial way: it can be contaminated. The same logic now applies to digital evidence. Just as a memory becomes less reliable the more it is re-exposed to suggestions and post-event information, a piece of digital evidence becomes harder to trust the further it travels from its original source without a verified chain of custody.
The increasing sophistication of deepfakes poses significant challenges for judges and attorneys in detecting and authenticating evidence, because generative AI creates content using two algorithms — one that generates content and one that evaluates its realism — creating a constant feedback loop that continuously improves the AI’s ability to produce convincing fakes.
The result, cognitively speaking, is an arms race that human perception is not equipped to win.
What Needs to Change
Experts say countermeasures must include certified video companies, chain of custody rules, and perhaps greater reliance on in-person testimony. At the judicial level, some legal scholars are pushing for a comprehensive framework for the evaluation and ongoing development of AI-powered legal tools, while other experts advise judges to consider the credibility of witnesses and order in-person appearances when necessary.
But psychological education for jurors may be just as urgent. Research has shown that jurors who lack accurate knowledge of how memory and perception work are far more likely to make flawed evidentiary judgments. The same will be true of AI literacy. A juror who understands, conceptually, that a video can be fabricated may approach digital evidence with the same healthy skepticism they’d apply to any other testimony — weighing it, rather than simply believing it.
We have spent decades building legal safeguards around the fallibility of human memory. We are now at the beginning of a longer, harder project: building safeguards around the fallibility of perception itself.