For years, the fear was that fakes would become good enough to deceive. The most insidious danger lies elsewhere: it is now enough for fakes to exist for the true to become contestable. When everyone knows that any image, voice or video can be fabricated, the liar no longer needs to produce a convincing fake: faced with damaging but authentic evidence, it is enough to say "it's a fake". Legal scholars have given this windfall a name: the "liar's dividend".
1 It only takes a word: "it's a fake"
The concept, and its founding paradox.
The liar's dividend
The fake first profits those who lie
The phrase comes from legal scholars Robert Chesney and Danielle Citron, in an article in the California Law Review (2019). Their idea: as the public learns that anything can be faked, those caught out can reject an authentic piece of evidence by declaring it fabricated. The paradox is cruel: the more we educate people about deepfakes, the more we strengthen the liar's argument. As they write, the liar's dividend "grows in proportion to the public's awareness of deepfakes". Raising awareness of the threat feeds the threat.
2 The effort asymmetry
The heart of the title: accusing and proving do not weigh the same.
The imbalance
An accusation takes three words, a proof takes three weeks
The mainspring of this dividend is an asymmetry of cost. Accusing is free, instant and understandable by everyone: "it's a fake" is said in a second and shared in a click. Proving authenticity, by contrast, is slow, costly and hard to read: it takes forensic expertise, specialized tools, technical vocabulary, and the verdict always arrives after the doubt has spread. The researcher Renée DiResta noted as early as 2018 that merely "pointing out that the technology exists" is enough to taint the integrity of what is real. And as the image expert Hany Farid reminds us, forensics runs behind the fake: it is always one step too late.
The cost of doubting
≈ 0
a single sentence, no skill required, instant spread.
The cost of proving
High
expertise, time, tools, and a verdict that arrives too late.
An older asymmetry
The liar's dividend is only a special case of a far more general law. The developer Alberto Brandolini formulated it in 2013, under the name "Brandolini's law": "The amount of energy needed to refute bullshit is an order of magnitude bigger than that needed to produce it." Long before him, Jonathan Swift had already observed: "Falsehood flies, and the truth comes limping after it." The cause is as much cognitive as technical: a false claim that is simple, fluent and immediately appealing is adopted without effort, whereas an exact truth, often nuanced and laborious, demands that we stop and think. Refuting will always cost more than asserting; and the fake keeps, over the true, the advantage of ease.
3 The dividend, measured
This is no theoretical fear: research has quantified it.
The test of the data
Crying fake really does pay
A study published in the American Political Science Review (Schiff and colleagues, via Cambridge), run on more than 15,000 participants, tested the mechanism: faced with unfavorable information, an official who cries disinformation increases the support given by their own backers and erodes trust in the media. The effect is robust when the accusation targets text or claims, more limited when it faces a video. The Brennan Center and other work converge: denying reality is a strategy that, on average, pays. The liar's dividend is not a metaphor; it is an observable, measurable behavior.
4 When the real proof is denied
Concrete cases where the authentic was branded a "fake".
From concept to fact
The "deepfake" defense enters the courtroom
The tactic is no longer hypothetical. In a dispute over a Tesla crash in "Autopilot" mode, the company's lawyers suggested, in 2023, that authentic video statements by Elon Musk might be deepfakes; the judge dismissed the argument, finding it troubling that an executive could escape his own words by invoking the possibility of a fake. In politics, authentic images or recordings have been contested as "doctored" to dodge accountability, from the United States (episodes tied to rallies or to the Capitol riot) to India (Tamil Nadu) and Gabon, where doubt over a presidential video fed a political crisis. The same tool, everywhere: denying the authentic by brandishing the specter of the fake.
The reversal
The irony is complete: the technology of the fake, supposed to serve deception, first serves denial. The most effective deepfake is not always the one that is made; it is sometimes the one that is invoked to reject an inconvenient piece of evidence. The threat lies not only in the fakes that circulate, but in the generalized suspicion they license.
5 Reality apathy
When we give up telling the true from the false.
Epistemic fatigue
The real danger: ceasing to seek the truth
The heaviest cost is not that we believe a fake, but that we end up believing nothing at all. The technologist Aviv Ovadya named this tipping point "reality apathy": overwhelmed by information whose authenticity they cannot verify, the public gives up sorting and retreats into its preferences. When "anything can be fake", the burden of proof shifts from the one who lies to the one who tells the truth: it now falls to the victim, the journalist, the judge, to prove the authentic. And a democracy that can no longer establish shared facts loses the ground on which justice, information and the vote all rest.
6 Signing the real
The technical answer: prove the origin, rather than hunt the fake.
Digital provenance
Affixing a signature at the source
Since detecting every fake is a losing race, another path is to certify the true at the source. This is "digital provenance": attaching to a piece of media, from the moment of its creation, a cryptographic certificate attesting where it comes from and how it has been modified. The open C2PA standard and its "Content Credentials" component bring together, in 2025, more than 5,000 members, from device makers to software publishers. Google has watermarked more than 10 billion pieces of content with its SynthID tool. And cameras now sign the image at the moment it is captured (Leica, Sony, Samsung). The idea is not to prove that a file is fake, but to prove that a file is, for its part, what it claims to be.
C2PA coalition (2025)
> 5,000
members adopting the "Content Credentials" standard.
SynthID watermarks
> 10 billion
pieces of content marked by Google's tool.
7 The limits of proof
Provenance shifts the burden; it does not abolish it.
What the signature does not say
Proving a signal is not proving the truth
Provenance is no silver bullet. Watermarks can be removed or altered: researchers, such as the team of Soheil Feizi at the University of Maryland, have shown that these marks can be erased or imitated. Above all, a signature attests to an origin, not to a truth: an image authentically captured can show a misleading scene. Sam Gregory, of the NGO WITNESS, puts it plainly: provenance "provides signals but does not prove truth". And a trap awaits: if we grow used to demanding a certificate, the absence of a certificate will end up passing for proof of a fake, even though the vast majority of real content carries none.
8 The panic factor
Let us be honest: the effect is real, but it has bounds.
The nuance
Do not overstate the threat
The liar's dividend is documented, but alarmism has its own dangers. Researchers at the Harvard Kennedy School Misinformation Review (Felix Simon, Sacha Altay, Hugo Mercier, 2023) stress that the effect owes less to technology than to partisanship: people believe, or reject, first according to their side. Fact-checking, when it intervenes, largely cancels out the gain of the one who cries fake. And we know few cases where a deepfake has, on its own, swung an election; the often-cited example of a dubious recording released just before an election in Slovakia (2023) remains disputed. Overstating the threat is itself a risk: by endlessly repeating that "nothing is reliable", we feed precisely the apathy we claim to fight.
The middle ground
The right stance is therefore neither panic nor denial. The liar's dividend is a real, measured effect, but a bounded one, mediated by pre-existing convictions and tempered by verification. Describing it accurately — without dramatizing or minimizing it — is already part of the response.
9 Proving the true
The asymmetry remains; it is up to us to learn to live with it.
What remains
Doubting will stay free, proving the real will stay dear
No technology will fully close the gap: it will always be easier to cast a doubt than to establish a proof. What we can do is more modest and more lasting: build provenance standards, extend signing at the source, train the public in discernment, and support a journalism and a justice capable of establishing facts. The expert Hany Farid sums up the stakes in a phrase: "if everything can be fake, then nothing has to be real". The real challenge is not only technical; it is epistemic. It is less about detecting fakes than about recovering the means to trust the true.
The compass
①
The liar's dividend. When anything can be faked, claiming "it's a fake" is enough to reject an authentic piece of evidence: the fake first profits those who lie.
②
An asymmetry of cost. Doubting is free and instant; proving authenticity is slow, costly and always behind. Digital provenance (C2PA, SynthID) shifts the burden without erasing it.
③
A real but bounded effect. Mediated by partisanship, tempered by verification. This sheet sheds light on an epistemic debate; it offers neither legal nor investment advice.
The thread of trust
This subject extends a worry already explored on the data side:
"Harvest now, decrypt later", where trust in encryption and in digital authenticity is also played out over the long term. On one side we doubt the true, on the other we fear for the secret: the same currency, trust, and the same erosion.
Read alongside: Harvest now, decrypt later, the other long-term side of digital trust. Reference: abbreviations & acronyms (AI, C2PA).