Big claims.
Better questions.

An accuracy estimate describes how strongly the available evidence supports a specific factual claim: 0–20 contradicted, 21–40 mostly unsupported, 41–60 mixed, 61–80 mostly supported, 81–100 strongly supported. It is an AI estimate, not a calibrated probability. Evidence confidence is reported separately. Opinions, predictions, inaccessible posts, and insufficient evidence should receive no score. A single-source answer can be shown without a numeric score. At least two domains with matching evidence quotations are required for a scored result; separate domains do not necessarily mean independent reporting. Follow links and check the original material.

Research uses Cloudflare Workers AI with free Full Wikipedia articles, Crossref research abstracts, and accessible references. Relevant passages and exact quotations are checked against the retrieved text. Coverage is limited and is not comprehensive web search. Only submitted content and the current follow-up context are sent. No chat history is saved by this app. Short-lived, salted address hashes are used to enforce network limits; raw addresses are not saved by the application. Cloudflare processes connection data for bot verification. Daily counts and temporary abuse-prevention records are stored. There is a site-wide limit of 40 attempts per UTC day, including follow-ups and research failures. Each network address has 20 attempts per day with a 30-second cooldown; shared Wi-Fi users share that limit. Another check in progress will show a retry message. Turnstile verification is required for every check. X posts may be inaccessible; paste their text when requested.