What about the lawyer who says, “I won’t use AI at all”? It’s an understandable reaction. The machines hallucinate—when allowed to. They confidently invent cases, quotations, and propositions. But refusing to use a powerful tool can be as foolish as trusting it blindly. Guided well, AI can spot patterns, generate possibilities, expose arguments, and dispatch routine work at astonishing speed. The sensible choice isn’t between embracing AI and shunning it. It’s to learn what it does well, where it goes wrong, and how to check its work.
That matters because AI is making legal information, preliminary research, and polished prose cheap and fast. In seconds, a lawyer can get a case summary, doctrinal overview, list of authorities, statutory comparison, or draft memorandum. Useful? Often. Authoritative? No. These are orientation tools. They can point toward issues, authorities, arguments, and counterarguments, but they don’t establish what a statute says, what a case held, whether an authority remains good law, or whether the record supports a factual assertion. All the talk about hallucinations can obscure this larger problem. An answer needn’t be fabricated to be misleading. It need only sound more certain than the evidence warrants.
That’s where the lawyer’s real work begins. Read the primary sources. Find the controlling cases. Look for the qualification, the adverse authority, the procedural wrinkle, the subsequent treatment. Check the current law. Test the facts against the record. A citator, annotated code, reliable treatise, official docket, and primary-source record remain indispensable because they provide provenance and an inspectable basis for legal propositions. AI can tell you that a court adopted a rule. Only the lawyer can determine what the court actually decided and whether that decision governs the client’s case. AI can produce an answer in seconds. The lawyer remains responsible for knowing whether it’s an answer worth giving.