Public discourse·intermediate

Liar's dividend

The existence of fakes lets anyone dismiss real evidence as fake.

Also known as Liar's dividend (Bobby Chesney & Danielle Citron, 2019)

When AI-generated images, deepfakes, and fabricated content become common enough, the default assumption shifts: every inconvenient real image can be dismissed as AI. The technology that lets anyone fabricate evidence also lets anyone deny genuine evidence by claiming it was fabricated. The cost of lying drops; the cost of proving truth rises.

Truth-adjacency

Truth-adjacent: the pattern's significance depends on whether the claim is true

Where it shows up

Public discourse

How it works — spans two mechanisms

Verification exhaustion Synthetic output and proxy drift

What to watch for

The phrases and tells that mark this pattern in the wild:

this is AIthis is fakethis is a deepfakeyou can't trust photos anymorecould be AI-generatedhow do we know this is real

How to recognize it

The claim is not “this is fake because of X artifact.” The claim is “this could be fake, therefore it is fake.” The possibility of fabrication is treated as proof of fabrication. The burden of proof shifts from the person making the denial to the evidence itself.

The Deceit question

Is there specific evidence that this particular image, video, or document is fabricated, or is the claim just that fabrication is possible in general?
What verification would satisfy the person making the denial? If the answer is “nothing,” the denial is the pattern, not a genuine inquiry.
Who benefits from the doubt? The liar’s dividend is most useful to the party that needs real evidence to be dismissed.

What it looks like when you’re wrong about it

You call “liar’s dividend” on someone who is genuinely asking whether an image is AI-generated because it has visible artifacts or no verifiable provenance. That’s legitimate verification, not the dividend. The pattern requires the dismissal to be unfalsifiable, no amount of verification satisfies the doubter, and the possibility of fakes is used as a permanent license to reject.

Spot the pattern

One of these two real scenarios is Liar's dividend. The other is a different pattern entirely. Which one is which?

What it feels like from the inside

Like trying to prove something real to someone who has decided in advance that nothing can be proven.
Like the ground rules of evidence shifting mid-argument, so that what counted as proof yesterday counts as suspicion today.
Like the existence of counterfeits being used to argue that all currency is fake.

How it starts

Real evidence of wrongdoing surfaces. The accused responds not with counter-evidence but with the claim that the evidence could be AI-generated, exploiting the public's awareness that deepfakes exist.

How it progresses

The default assumption shifts: every inconvenient image, video, or recording is met with suspicion before it is met with analysis.
The burden of proof inverts. Fact-checkers must prove authenticity rather than the accused proving fabrication.
The doubt becomes unfalsifiable. No amount of verification satisfies a doubter who treats the possibility of fabrication as proof of fabrication.
The cost of lying drops and the cost of proving truth rises, because every piece of real evidence now carries a built-in escape hatch.

Common signs

Real evidence dismissed with 'this could be AI' and no specific artifacts cited to support the fabrication claim.
The accused demanding verification standards that no amount of evidence could satisfy.
Public discourse where the possibility of deepfakes replaces actual analysis of specific media.
A pattern where only inconvenient evidence is questioned, while convenient evidence is accepted without scrutiny.

Why it's hard to leave

The doubt is structurally reasonable because deepfakes do exist. Dismissing the concern entirely would be naive, but accepting it uncritically gives every bad actor a permanent get-out-of-evidence-free card.

Do this now

Ask for specific evidence of fabrication. 'This could be AI' is not an analysis. Visible artifacts, metadata anomalies, or provenance gaps are.
Check what verification would satisfy the doubter. If the answer is 'nothing,' the denial is the strategy, not a genuine inquiry.
Notice the selectivity. If only inconvenient evidence is questioned while convenient evidence is accepted, the skepticism is a tool, not a principle.

What people realize later

Later, people realize the possibility of fabrication was being used as a substitute for evidence of fabrication. The existence of fakes had become a license to reject anything real that was inconvenient, and the doubt was the product, not the inquiry.

Recognized this in public information?

This pattern in the wild

Field notes where this pattern was identified:

Trump was right about the fakes(Fabricated)
The AI image that proved what you already believed(Fabricated)

Misuse Guardrails

How this pattern gets misused

A bad actor dismisses genuine evidence of wrongdoing by claiming 'that could be AI' without providing any evidence of fabrication. The vocabulary becomes a blanket denial tool, the existence of fakes becomes a license to reject anything inconvenient.

What it looks like when you're wrong about it

It's only the liar's dividend when real evidence is dismissed using the mere possibility of fabrication, without evidence that the specific item is fabricated. Questioning whether something is AI-generated is legitimate when there are actual artifacts or provenance gaps. The pattern requires the dismissal to be unfalsifiable, no amount of verification satisfies the doubter.

Not sure? Describe the situation to someone outside it. If they do not see the pattern, pause before you name it.

Related Patterns

Source obfuscation

You trusted it because it sounded official. 'Official' was the costume, not the credential.

Authority laundering

You stopped asking for the evidence because the phrase sounded settled. 'Experts agree' was the whole argument. The authority was borrowed, never shown.

Agentic coordination

Nobody wrote the post. Nobody had to. Somebody only had to give it a goal.

Commonly stacks with

Misjudgments compound rather than act alone. This pattern is often deployed alongside:

Manufactured evidence

The crash was real. The proof of it wasn't. Somebody built the fire, then filmed it catching.

The name is designed to spread. The hook is designed to stick. If you recognized something, share the name.

Seen a real example of liar's dividend? Suggest it for the Register →

Share on X Share on Threads