ThinkableWhat is this?

Technology & Society

The Liar's Dividend: Why the Real Danger of Deepfakes Is What They Do to Real Things

The most corrosive thing deepfakes have done is not make us believe fake videos — it is give us a reason to disbelieve real ones.

The Idea

For most of human history, a video was about as close to proof as evidence got. Courts used it. Journalists relied on it. People trusted their eyes. Deepfakes have destabilised all of that — but not quite in the way most people imagine. The obvious fear is that someone fakes a video and you believe it. That happens, and the harms are real. But there is a subtler, arguably more dangerous effect that legal scholars Bobby Chesney and Danielle Citron named the liar's dividend back in 2019: once people know that convincing fakes are possible, any inconvenient real recording can be dismissed as fabricated. No deepfake is even required. The mere existence of the technology is enough to cast doubt. As one analysis put it, the liar's dividend is the advantage a dishonest actor collects from public awareness of deepfakes — the better educated the audience is about fakery, the easier it becomes to wave away something genuine. This double-edged quality is what makes the problem so structurally hard to fix. Raise public awareness of deepfakes, and you simultaneously raise the plausibility of denying real evidence. Human visual inspection is now essentially unreliable — research suggests people achieve only roughly fifty to sixty percent accuracy when asked to identify modern deepfakes, little better than a coin flip. Automated detectors are more accurate, but they are locked in an arms race with generation tools, and both sides keep improving. The technology to create convincing synthetic media has outpaced the social and institutional norms we have built around trusting our senses.

In the World

In January 2024, a finance worker at the Hong Kong office of Arup — the British engineering firm behind buildings like the Sydney Opera House — received a message that appeared to come from the company's UK-based chief financial officer. It described a confidential transaction and invited him to a video conference to discuss it. He joined. On the call he saw the CFO and several senior colleagues. He heard their voices. He recognised their faces. He authorised fifteen transfers across five bank accounts, totalling roughly twenty-five million US dollars. Only when he contacted head office afterwards was the fraud discovered. Hong Kong police later confirmed that every other person on that call had been a deepfake, constructed from publicly available recordings of the people being impersonated. No internal systems were hacked. The payment process simply trusted what it saw. The Arup case sits at one end of the harm spectrum. At the other sits the liar's dividend in action. During the 2026 conflict involving Iran, the proliferation of AI-generated videos led to real footage — including a video posted by Israeli Prime Minister Benjamin Netanyahu — being widely dismissed online as synthetic. A researcher does not need to prove a video is fake to neutralise it; raising the question is enough. In the same week in late October and early November 2026, a Russian influence operation identified as Storm-1516 re-voiced and lip-synced real local news broadcasts from US television stations to deliver fabricated allegations against Senate candidates — borrowing a real newsroom's credibility while corrupting its content. The candidates won their races, but the harm to real people whose words were put in their mouths was not undone by the election result.

Why It Matters

The practical upshot is that our epistemic infrastructure — the shared understanding of what counts as evidence — is under pressure from two directions simultaneously. Fakes circulate as real. And real things get labelled fake. This matters for how you consume information. The instinct to become a sceptic about everything you see is understandable, but research suggests it comes with a cost: interventions that improve people's ability to spot deepfakes also reduce their ability to correctly identify genuine media. Healthy scepticism, pushed too far, tips into blanket distrust, which is exactly what bad actors are counting on. What does help is shifting from asking 'does this look real?' to asking 'where did this come from?' Provenance — the documented chain of custody from capture to publication — is becoming the new basis for media trust. Initiatives like C2PA (the Coalition for Content Provenance and Authenticity) embed cryptographic records directly into files, so that tampering breaks a verifiable seal. That is not a perfect solution, but it points toward the right frame: the question is not whether your eyes are reliable, but whether the systems around the content are. In a world where seeing is no longer believing, we need better reasons to trust than the evidence of our senses alone.

A Question to Ponder

If you can no longer trust a video as proof that something happened, what standard of evidence would you actually find convincing — and who gets to set that standard?

One reader's Thinkable for Tech Tuesday — a random draw from the library.

Start learning with Thinkable
Get your own draw, every day.Start free