Is your AI making things up?

A hallucination is when an AI model states something false or unverifiable with total confidence — a made-up statistic, a fake citation, a wrong date, or an event that never happened. Because the text reads fluently, these errors are easy to miss. This tool surfaces them so you can fix or remove them before publishing.

We split the text into individual claims, then assess each one for verifiability, internal consistency, and the linguistic patterns that correlate with fabrication (false precision, hedging, invented specifics). On Pro, each supportable claim is additionally matched against live web sources.

No automated checker is perfect — treat the result as a prioritized list of what to double-check, not a final ruling. It dramatically narrows where to focus your fact-checking, especially on long answers, but a human should confirm anything high-stakes.

Any of them. Paste output from ChatGPT, Claude, Gemini, Perplexity, Llama, or any other LLM — the analysis is on the text itself, so the source model doesn't matter.

No. Free checks are analyzed transiently and never used to train our models. Pro checks can be saved to your private history if you choose, and deleted any time.

Because a language model is built to predict fluent-sounding text, not to tell the truth. When it doesn't actually know something — a precise statistic, a source, an exact date — it fills the gap with a plausible guess instead of saying "I don't know." Rare facts and specific citations are where models invent the most, which is why those are the first things we check.

Not reliably. The same model that invented a fake citation will usually defend it when you ask, because it has no memory of what's real — only what sounds right. An independent check that verifies claims against outside sources catches what the model can't see in itself. Our AI fact-checker works the same way for general claims.

Paste the text and we pull out every DOI, arXiv ID, PubMed ID, ISBN and link, then look each one up in the real database — CrossRef, PubMed, arXiv and Open Library — plus a live check on any URL. A reference that doesn't resolve is flagged as fabricated, backed by the actual lookup rather than a guess. Need to format real citations instead? Use our citation generator.

An AI detector estimates whether text was written by an AI. A hallucination detector checks whether the facts are true. They're different jobs: human-written text can be full of hallucinations, and AI-written text can be perfectly accurate. Use this tool when you care about accuracy, not authorship.