About Claim Classifier

Made by the same person who writes Philosophy and Beyond.

Built during MiSo Camp Tokyo 2026, a program of the Henkaku Center for Radical Transformation (Chiba Tech, Japan). The camp was designed and run by Noah Raford and Igor Schwarzmann.

The goal of this tool is not just to hand you a verdict, but also to help you learn to classify claims yourself: to spot the difference between a claim that is simply false and one that is a lie, between something genuinely unverifiable and something a bit of digging could resolve, between an opinion dressed up as fact and a fact people wish were opinion.

What the categories mean

Factual

Verifiably true or false as a matter of record. The evidence settles it.

Speculative

A projection or inference reasoned from known facts — a forecast, an estimate, a likely explanation. Not yet settled because it concerns what is probable or still to come, rather than because the evidence is blocked.

Opinion

A value judgment or belief that no evidence could settle. Covers both "X is better than Y" and metaphysical claims like the existence of God — assertions about the world, but not ones any observation could decide.

Lie

The speaker knows the truth and says the opposite, with intent to deceive.

Lapse

The speaker believes it — but it's wrong. No intent to deceive, just a genuine mistake.

Bullshit

The speaker isn't oriented toward truth at all. The claim's function is persuasion or effect, not accuracy. This is different from lying: a liar cares about the truth enough to hide it; a bullshitter doesn't care either way.

Unverifiable

Checkable in principle, but the available evidence doesn't settle it — single-source reporting, inherently opaque activity, or credible sources that directly conflict with no way to adjudicate. The evidence is blocked or missing; its absence tells you nothing either way.

Unsupported

The claim predicts evidence that should exist and be findable — people competent to find it have looked, and it isn't there. Here the absence is itself informative. This is the difference between "we can't tell" and "we looked where it should be, and found nothing," which is why it sits apart from both Unverifiable and simple falsehood.

Not a checkable claim

The submitted text isn't a single evaluable assertion: a headline bundling several accusations, a question, a topic label. Nothing here can be true or false in isolation.

Claim Classifier is not affiliated with any political party, media outlet, or advertiser. Every classification is generated automatically by an AI model — see our Terms of Use for what that means in practice, including where it can go wrong.

The person behind this project has no influence over what gets submitted or how any individual claim is classified — every request runs through the same automated process, with no manual editing of results. The public archive is the one exception: it is manually curated, meaning only a subset of submitted claims are selected for publication, specifically to avoid ethical issues such as defamation or endangering a private individual. Curating the archive is a decision about what to publish, not about what a classification says.

Questions, feedback, or a claim you'd like reviewed? Reach out via Philosophy and Beyond.