Ask a question, brief a case, find the exact quote, grade your practice exam, or check if a case is still good law. Multiple AIs check each other, and every citation is verified against real case law — so you never rely on one that doesn’t exist, or doesn’t say what it’s cited for.
Ask a chatbot a legal question and it will often cite a case that sounds perfect — the right court, a plausible reporter cite, a clean holding — that does not exist. Students have turned in briefs built on invented authority; lawyers have been sanctioned for filing it.
“The Court is presented with an unprecedented circumstance… six of the submitted cases appear to be bogus judicial decisions with bogus quotes and bogus internal citations.”
— a federal judge, sanctioning lawyers who filed AI-hallucinated citations (Mata v. Avianca, 2023)A real example: a common FDCPA question. Two models cite good law; one invents a case — and the ledger catches it on the spot.
Each tool runs on the same trust engine: multiple AIs, cross-checked, with every citation verified against real law.
Put a legal question to Claude, ChatGPT & Grok. A moderator cross-examines them, surfaces the minority view when they split, and every case they cite is verified and attributed.
Give it your proposition; it finds the controlling case and the verbatim passage that supports it — confirmed by multiple AIs. No real support, no quote.
NewEnter a name or citation and get a clean brief — facts, issue, rule, holding, dissent, significance — reviewed by a second AI, with every citation checked. Export to PDF.
Paste a brief, memo, or paper. Each citation comes back verified, fabricated, or misattributed — before it costs you a grade or a sanction.
A panel of AIs grades your practice exam on IRAC and averages the scores — and any case you cite is verified, so a fabricated cite is caught, not rewarded.
NewBefore you rely on a case, check whether it’s been overruled, limited, or criticized. A panel checks later authority and tells you — conservatively — when to double-check.
NewA single model sounds just as confident when it’s wrong as when it’s right. Legal Consensus runs your work past multiple models, has a moderator cross-examine them, and then verifies every citation against a real database of decisions. When the models disagree, you see the split — not a false consensus. And when a citation can’t be backed by a real opinion, it’s flagged, not smoothed over.
Ask the panel, brief a case, find a pincite, grade a practice answer, check a case’s treatment, or paste a draft to check.
Multiple models respond; a moderator cross-examines them, surfaces the minority view, and flags where they overreach or disagree.
Each case is checked against the CourtListener/Caselaw corpus and shown as verified, fabricated, or misattributed — attributed to whoever cited it.
Never hand in a brief that cites a case which doesn’t exist — or doesn’t hold what you say it does. Find the exact pincite before you file.
Grade your practice answers against an IRAC rubric with a panel of AIs, and get verified answers to the doctrine you’re shaky on.
Cite-check a source set fast during source-and-cite, and confirm a case is still good law before you build on it.
Save and organize everything by course, reopen a question, export to PDF, and share a verified answer with a link.
A credit = one citation checked. Draft checks draw one per citation; panel questions, briefs, pincites, graded answers, and treatment checks each draw a small flat amount. Prices in USD.
Two ways. First, it’s more than one AI — several models answer, a moderator cross-examines them, and you see where they disagree instead of one confident voice. Second, and most important, every citation is verified against a real database of decisions — so the fabricated cases that get students and lawyers in trouble are caught, not repeated.
Yes — the Still good law? tool has a panel check whether a case has been overruled, limited, or criticized, and it errs on the side of caution when the models disagree. Treat it as a fast signal, then confirm anything you’ll rely on with Shepard’s or KeyCite.
No. It’s a research aid for students — not legal advice, and not a substitute for Shepard’s or KeyCite. It helps you catch citations that don’t exist or don’t support your point before you rely on them.
Existence is checked against the CourtListener/Caselaw corpus — a real database of decisions. Whether a case supports the point is judged by an AI that must quote the opinion to back its call, cross-checked a second time (and, with more than one model configured, confirmed by a second AI). When it isn’t sure, it abstains rather than wrongly accusing a real case.
No — anyone can use the Free tier and buy credit packs. A verified .edu email is what unlocks the $8/mo Student plan.
Yes. Your drafts and saved checks are private and are never published. Only a panel answer you explicitly choose to publish becomes public — and even then it’s just the answer, reviewed first.
Claude, ChatGPT, and Grok, with an AI moderator that cross-examines them. What makes the difference isn’t which model — it’s that every case any of them cites gets checked against real law.
Ask your first question or check a draft in under a minute — free, no card.
Try Legal Consensus →