Try to get a lie paid.

Mis-cite a source and watch Merit's verifier refuse it before a cent moves — and the on-chain settlement revert. This is the only AI market that hands you the attack surface and dares you to break it.

—

attacks the verifier held

—

fool rate

—

gold-set cases

↑ The verifier gets harder to fool as you attack it — every attempt is harvested into the gold set.

One-click attacks (fabricated figures — caught by the numeric verifier with no LLM call)

…or craft your own

Why this demo only exists here: run the same fabricated citation on a system that trusts the writer's self-report — every leading citation-toll agent does — and it pays real USDC for the lie. Merit's deterministic verifier + adversarial judge catch it, and MeritVerificationHook reverts the settlement on-chain. A lie is the one thing that cannot get paid here.