How to verify AI legal research before you cite it

Generative research can feel magical: a full paragraph of law, with case names and paragraph numbers, appears in seconds. But the same speed makes verification non-negotiable. Before any AI-sourced authority goes into a submission, run it through a short checklist.

1. Read the original source

Open the judgment, statute or regulation the AI points to. Do not rely on a summary. Check that the quote exists, that the paragraph number is correct, and that the ratio has not been stretched beyond what the court actually decided.

2. Check the date and jurisdiction

A good authority from the wrong jurisdiction is still the wrong authority. Confirm the court, the year and whether the decision has been followed, distinguished or overruled. If your matter is outside the United States, make sure the source is from the relevant jurisdiction rather than a familiar US case.

3. Match the authority to the question

An authority is useful only when it answers the precise point you are making. Ask whether the cited case deals with the same procedural posture, the same cause of action and the same relief sought. If it does not, keep searching.

4. Look for contradictory authority

A one-sided list is a weak list. Search for cases that cut the other way and explain why they do not decide the point in your favour. Courts expect an honest account of the landscape.

5. Record your sources in the file

Paste links, excerpts and notes into a shared research log so the rest of the team — and the partner reviewing the draft — can see what was checked and when.

AI legal research is a starting point, not an endpoint. The best litigators treat it as a fast first draft that still demands the same professional verification they would apply to a junior associate's memo.