The Liar’s Dividend: How Deepfakes Changed the Burden of Proof
A website, or taking your word for it, is no longer enough to establish an organization as trustworthy to the American public.
-By Laura Rodriguez-Carbone, MPA, PMP
In 2016, Elon Musk was recorded saying that a Tesla on Autopilot could drive itself more safely than a person. Seven years later, the family of Walter Huang, an Apple engineer killed in 2018 when his Model X crashed on a California freeway with Autopilot engaged, wanted to question Musk about that claim. Tesla's lawyers responded with a new kind of argument. Musk is famous, they said, and famous people are frequent targets of deepfakes, so the court couldn't be sure the recording was real.
Judge Evette Pennypacker called that position "deeply troubling." In her tentative ruling, she wrote that Tesla's argument amounted to saying that because Musk "is famous and might be more of a target for deep fakes, his public statements are immune." She ordered him to sit for a deposition. The argument failed in that courtroom, but it showed where the damage from AI-generated media lands first. Fake video threatens the public, and the mere possibility of fake video gives powerful people a way out of the real record.
The Liar's Dividend
For most of modern history, institutions could settle an argument by pointing to evidence almost no one could forge. A video showed what a governor said at the podium, and a signed memo showed what an agency had promised. Generative AI broke that arrangement, and it did so in two directions at once.
The first direction is familiar. Fake content can now look real, and a fabricated clip of a public official can reach millions of people before anyone catches it. The second direction does more lasting harm. Law professors Robert Chesney and Danielle Citron named it in a 2019 California Law Review article, "Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security." They called it the liar's dividend, the benefit that flows to anyone caught on tape once the public knows tape can be faked. Translated into practice, "that video is fake" has become a defense that costs the speaker nothing. The lie never has to persuade a majority, because hesitation alone can stall accountability until public attention moves elsewhere.
Most People Can't Tell the Difference
In February 2025, the identity-verification company iProov tested 2,000 consumers in the United States and the United Kingdom on a mix of real and AI-generated images and videos. Only 0.1% of participants correctly identified every one. More than 60% remained confident in their detection skills regardless of whether their answers were correct. That gap between confidence and accuracy explains how fakes travel, since people share what they feel sure about and the study shows that feeling is a poor guide to the truth.
The public already senses the shift. In a 2023 survey of 2,000 U.S. adults by Pindrop and Voicebot.ai, more than 90% expressed at least some worry about deepfakes and voice cloning. In the iProov study, 49% said learning about deepfakes made them trust social media less. Nadia Naffi, an education researcher at Université Laval, argued in a 2025 essay for UNESCO, "Deepfakes and the crisis of knowing," that deepfakes weaken the methods societies use to establish what is true. For anyone whose work depends on being believed, that means accuracy alone no longer protects them.
What This Means for Communicators
Accuracy has always been a communicator's strongest asset. It now also has to be checkable by a stranger who has no reason to trust you. That standard calls for provenance, meaning a direct link to the primary source or a dated public record. A claim that asks readers to take your word for it asks them for faith, and faith is the resource institutions now have least of.
Conduct also outweighs messaging when trust runs low. Anyone can now fabricate a statement in your name, so the public record of what your institution did, and when, becomes its strongest defense. People believe institutions whose actions are documented in public and whose statements match those actions.
No organization can answer every fabricated clip, and each rebuttal risks pushing a fake in front of people who never saw it. Choose the fakes that could change a real decision or reach a large audience, and answer those with documents and dates. Leave the outrage out of it.
Trusted messengers and plain language matter more than they did before AI video arrived. People believe the people they know over the institutions they've learned to doubt, and no one trusts a statement they can't understand.
The Two-Click Test
Choose the most important claim your institution has made or plans to make, the one that matters most to the people you serve. Imagine a critic calls it a fabrication. Then ask how a skeptical reader could confirm it within two clicks. A link to your own website fails the test, because the skeptic already doubts you.
Build a proof page for that claim before anyone challenges it. The page should hold the primary source, dated records, and independent news coverage that confirms the facts, all in public view. Then make the practice routine and attach the same kind of proof to every statement that matters.
Detection tools will keep improving, and fakes will keep improving with them. The institutions that stay believable will be the ones whose record a is transparent and easy to find.
Publish your proof before anyone asks for it, and put it where your harshest critic can find it in two clicks.
Laura Rodriguez-Carbone was a federal public servant across 7 agencies including CMS, NIH, FDA, and FEMA before being purged in 2025. She ran for Congress in Ohio's 7th District in 2026. She founded Explainer Civic Strategies, a Lakewood, Ohio, firm that helps public institutions earn and keep the public's trust. Learn more at explainerconsulting.com.
