Unverifiable
Last updated: July 26, 2026
Unverifiable means a verifier found no authoritative, independent source to confirm or refute a claim. It is a real verdict, not a failure: it flags the claim for human review instead of guessing. Marking a claim unverifiable is what protects you from a false 'supported' when the evidence simply isn't there.
What does 'unverifiable' mean?
Unverifiable is the verdict a fact-checker gives when it cannot find an authoritative, independent source that confirms or refutes a claim. Not 'this is false' — 'we could not check this'. The two are very different, and conflating them is a common mistake.
A claim can be unverifiable for good reasons: it concerns private or non-public facts, it is too recent for sources to exist, it is too niche for any authority to cover, or the available sources are not trustworthy enough to rely on. In all these cases the honest answer is to decline, not to guess.
Unverifiable is one of the four verdicts, alongside supported, contradicted, and disputed.
Why 'couldn't check' is a real, useful verdict
The temptation for any verifier is to force a supported-or-contradicted answer on every claim. That is exactly the behaviour that makes checkers untrustworthy. When evidence is absent, forcing a verdict means guessing — and a guess dressed as a verdict is worse than no verdict.
Returning unverifiable instead does two things. It keeps the false-alarm rate low, because the system never flags a true claim as false just to avoid saying 'I don't know'. And it hands you an honest signal: a human needs to look at this. That is far more useful than false confidence in either direction.
This is the same discipline that makes a model saying 'I don't know' more trustworthy than one that confabulates. Declining to guess is a feature. See how it works.
A worked example
An AI answer states: "The startup's internal Q3 revenue was $4.2 million." There is no public filing, press release, or database that reports this. A dishonest checker might mark it contradicted (there's no evidence, so call it false) or supported (it sounds specific). Both are wrong.
The honest verdict is unverifiable: no authoritative independent source exists to confirm or refute a private figure. The claim is flagged for a human who can check the actual books — not silently passed, and not falsely condemned. Try it on the verifier.
Frequently asked questions
Does 'unverifiable' mean the claim is false?
No. Unverifiable means no authoritative independent source was found to confirm or refute it — the claim might be true, false, or unknowable from public sources. It is explicitly not contradicted, which requires a source that refutes the claim. Unverifiable is a request for human review, not a judgment of falsity.
Why not just guess when there's no source?
Because a guess dressed as a verdict destroys trust. Forcing supported-or-contradicted on unsupported claims raises the false-alarm rate and makes every verdict less reliable. Returning unverifiable keeps the checker honest and tells you precisely which claims still need a human. See how it works.
What kinds of claims tend to be unverifiable?
Private or non-public facts (internal figures, off-the-record statements), very recent events with no sources yet, highly niche topics no authority covers, and claims where the only available sources aren't trustworthy enough to rely on. In each case declining is more honest than guessing.
Verify what your AI just told you.
Paste any AI answer and Mephistopheles checks each claim against independent sources — no sign-up to try.