liar's dividend
The benefit the liar gains from being able to easily label a real recording as a fake in the era of deepfakes.
In plain terms
When we think of deepfakes, we usually think of someone saying something in a fake video that they never said. The liar's dividend is the reverse of this: someone actually said or did something, there is a recording of it, yet they claim, "this is just an AI-made forgery."
It is like when word spreads in a city that there are many counterfeit banknotes. From then on, real money is viewed with suspicion too, and anyone who does not want to accept a banknote can easily say it must be fake. The mere existence of counterfeits devalues the real thing.
It is called a 'dividend' because the liar has to do nothing to earn it. They do not produce the deepfakes; they simply harvest the benefits of general distrust. The more we hear that anything can be faked, the more credible their denial sounds.
How it works
- Generative AI now allows cheap and fast creation of lifelike fake images, audio, and video, and the public knows this.
- Because of this, people will become more cautious: they will no longer automatically believe what they see or hear.
- A real but embarrassing recording surfaces, such as a leaked audio file or video.
- The subject does not explain but states that the recording is fake and AI-generated. They do not need to prove this; planting doubt is sufficient.
- Part of the audience becomes uncertain, supporters feel confirmed, and accountability is delayed or absent because verifying credibility is slow and requires expertise.
An everyday example
A recording leaks in which a public figure says embarrassing things, and the next day they state that their voice was forged with artificial intelligence. In your news feed, you only see two conflicting claims and cannot decide whom to believe. The liar's dividend is precisely this uncertainty.
Who is affected
A common misconception
Limits and risks
- If someone calls a recording a forgery, that is not proof: check what independent experts, multiple sources, and the original publisher of the recording say.
- Deepfake detection programs are not infallible, so a machine verdict of 'real' or 'fake' alone does not settle the dispute.
- The phenomenon is dangerous in reverse: if you doubt everything, you play into the hands of liars just as much as if you believe everything.
- Sometimes the recording is genuinely fake, so denial is not always a lie. Only verification can resolve the question.
Similar, but not the same
- deepfake – A deepfake is the fake recording itself; the liar's dividend is the benefit arising from the ability to dismiss the real as fake.
- disinformation – Disinformation is a broader concept: it covers every deliberately spread falsehood. The liar's dividend is a specific case of it, where the false claim is precisely that a genuine recording is fake.
- fake news – Fake news sells a fabricated story as truth, whereas here an actual event is being presented as fabricated.
Related concepts
disinformation, informational trust, MI-etika
–
deepfake, synthetic media, watermarking, content authentication, media literacy
A brief history
The term was introduced by two American law professors, Robert (Bobby) Chesney and Danielle Citron, in a study on the dangers of deepfakes, which appeared as a manuscript in 2018 and was published in the California Law Review in 2019.
Where things stand
By the end of 2026, the phenomenon is no longer theoretical: in political scandals and court cases, a recurring defense is that the recording was made by AI. Countermeasures, such as markings verifying content origin and the transparency requirements of the EU AI Act, are only now being implemented.
Also known as
liar's dividend · the liar's dividend · the liar's dividend · liar's dividend · liar's dividend · liar’s dividend · the liar's dividend · deepfake defense
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.