For Lawyerss and Legal Professionals

Cognitive Sovereignty Checklist for Lawyerss

About 20 minutes Last reviewed March 2026

AI legal research tools like Harvey, Lexis+ AI, and Westlaw AI hallucinate citations and misstate holdings. Contract drafting tools embed hidden assumptions into templates you inherit without question. Your professional licence depends on your judgement, not on the speed of your AI output.

Cognitive sovereignty insight for Lawyers: a typographic card from Steve Raju

These are suggestions. Take what fits, leave the rest.

Download printable PDF

0 / 18 applicable

Tap once to check, again to mark N/A, again to reset.

Before you use AI for legal research

Define what you will verify before citing any case or statutebeginner

Before starting research in Harvey AI or ChatGPT, write down your rule: will you always check primary sources? Only for key holdings? Never skip verification for cases cited in your opinion letters. Your client's case depends on this boundary.

Test your AI tool on cases you already know wellbeginner

Run three to five research queries on established law you understand. Does Westlaw AI's summary match what you know? Does it cite cases correctly? If your tool fails on familiar law, you cannot trust it on unfamiliar territory.

Separate AI summaries from your own case analysisintermediate

When Casetext summarises a judgment, read the full text yourself before adopting that summary. AI often compresses holdings in ways that hide risk or miss nuance. Your risk analysis must be your own, not borrowed from the AI's compression.

Check every citation the AI provides against the official reporterbeginner

If Harvey suggests a case name, citation, and holding, verify it took those details from a real judgement. Hallucinated citations look real and read smoothly. A five minute check prevents a professional negligence claim.

Identify which jurisdiction rules your search strategybeginner

Before asking ChatGPT or Lexis+ AI to research a legal question, tell the tool exactly which court you are arguing in. AI will mix English common law, US law, and statutory schemes unless you constrain it hard. Vague searches produce plausible nonsense.

Document the research you did not dointermediate

If you used AI to research one strand of law and then stopped, write down why. If you relied on the AI to find counterarguments and it found none, note that. Your file must show you did the thinking, not just ran the tool.

Cross-check AI research against Westlaw or Lexis without AIadvanced

Run the same research query without AI on one of your important cases. Compare results. If the non-AI search finds key cases the AI missed, you now know the limits of your tool on that topic.

When you use AI to draft contracts or advice

Read any template the AI generates as if someone else wrote itbeginner

AI contract templates carry assumptions about risk, liability, and party intent. Before you copy clauses into your draft, ask: why is this term here? What does it do? Who does it protect? Many AI templates privilege one party without saying so.

Identify which terms the AI drafted and which you changedbeginner

Use track changes or a separate note to mark every clause you altered or deleted from the AI template. If a problem arises later, you need to show the court you made deliberate choices, not just used an AI template verbatim.

Test the contract against your specific client's facts before you send itintermediate

An AI contract template covers general cases. Your client's transaction is specific. Walk through each clause with your client's facts in mind. Which clauses protect your client? Which expose them? The AI does not know your client's business.

Spot assumptions the AI baked into the draftintermediate

AI often assumes low-stakes transactions or standard commercial terms. If your client's deal is unusual, the AI will not know. Look for language about payment terms, liability caps, and dispute resolution. Ask yourself: did the AI assume facts that are not true here?

Require yourself to draft one section without AI assistanceadvanced

Before using ChatGPT or Harvey to draft a clause, write one version yourself first. Then compare it to the AI version. You find out quickly whether the tool is doing your thinking for you, or whether it is just faster at a job you already own.

Write your own risk summary before reading the AI's risk summaryadvanced

If you use Westlaw AI or Casetext to summarise contract risks, write your own risk list first. Then check it against the AI's list. What did you catch that the AI missed? What did the AI flag that you did not? The comparison teaches you where your judgement differs from the machine.

How to protect your judgement over time

Build a weekly habit of research without AIbeginner

One hour per week, do legal research the old way. Read cases in the reporter. Walk through a statute word by word. Use the index in Halsbury's. This keeps your foundational research skills sharp and reminds you what AI skips.

Write a memo on why you rejected an AI suggestionintermediate

When an AI tool suggests a case or clause that you decide not to use, write one paragraph explaining your reasoning. This habit sharpens your judgement and creates a record of your independent thinking.

Review a junior lawyer's AI-assisted work by asking them to defend itintermediate

When a junior lawyer brings you a draft researched with Harvey or Westlaw AI, ask them to explain the three weakest points in their research. If they cannot articulate risk, they did not do the thinking. Push back until they own the work.

Keep a list of cases where AI research failedintermediate

Note every time you caught an AI tool misstate a case, miss a key holding, or confuse two different lines of law. Over three months you will see patterns in where your tools fail. That map tells you when to trust the AI and when to check its work manually.

Require yourself to articulate the opposite positionadvanced

After an AI tool has argued one side of a legal question, spend ten minutes building the strongest counter-argument you can. If the AI did not mention this counter-argument, note it. Your job is to see risk the machine missed.

Five things worth remembering

When a case comes across your desk, ask yourself what you would have advised before you asked the AI. Then ask the AI. Compare the two answers. Over time you will learn exactly where your tool blinds you.
Never cite a case to a court until you have read it yourself. A hallucinated citation destroys your credibility and your client's case. The five minutes it takes to verify is not a cost, it is insurance.
If your junior lawyers start saying 'the AI says the law is' instead of 'I think the law is', you have a problem. Make them defend their positions as if you disagreed. Keep their judgement alive.
When you draft from an AI template, write in the margin next to each clause: 'AI wrote this' or 'I changed this'. This habit forces you to read the template critically and protects you if the clause later causes problems.
Once a month, run the same research query two ways: with AI and without. Compare results. This teaches you what your tool finds that humans miss, and what humans catch that the tool overlooks. You need both.

Prompt Pack

Paste any of these into Claude or ChatGPT to pressure-test your own judgment. They work best when you respond honestly before reading the AI reply.

Test your grip on AI-summarised case law

I have just used an AI tool to summarise a judgment. Ask me questions that would reveal whether I understand the actual holding, its limitations, and how it would likely be applied, or whether I am relying on the summary without the underlying reasoning.

Find what opposing counsel might exploit

I am preparing [describe matter]. I have done AI-assisted legal research. Act as competent opposing counsel and identify three arguments or authorities I have probably missed that you would use against my position.

Check your drafting instincts

I want to test my drafting judgment. Give me a contract clause to review and ask me what I would change and why, before offering any AI suggestions or analysis.

Pressure-test your client advice

I am about to advise a client on [describe situation]. Act as a senior partner reviewing my advice. Challenge the reasoning, flag the assumptions I have not tested, and identify where I might be creating professional risk.

Audit your AI reliance on a matter

Walk me through a series of questions about how I have used AI tools on my current matter. Help me identify which parts of my analysis are genuinely mine and which I have effectively outsourced, then tell me where the gaps are.

Reading List

Five books that give this topic the depth it deserves. Each one is genuinely worth reading, not just citing.

Thinking, Fast and Slow

Daniel Kahneman

Legal reasoning is built on the same cognitive machinery that produces systematic bias. Understanding how your judgment actually works is essential before you let AI assist it.

The Checklist Manifesto

Atul Gawande

The best examination of how structured process improves high-stakes judgment. Directly applicable to how you build AI verification habits into your legal workflow.

Weapons of Math Destruction

Cathy O'Neil

How algorithmic systems embed and amplify bias. Relevant for any lawyer working in areas where AI tools are being used in litigation, compliance, or regulatory decisions.

The Shallows

Nicholas Carr

Sustained reading of complex material builds the analytical capacity that deep legal reasoning requires. This book examines what digital habits do to that capacity.

Cognitive Sovereignty

Steve Raju

A practical framework for protecting your professional judgment as AI tools become standard in legal practice.

Questions to ask yourself

Use these before your next AI-assisted decision. Honest answers are more useful than comfortable ones.

When did I last verify a legal citation myself rather than trusting an AI tool?
Is my ability to construct a legal argument from scratch as strong as it was before I started using AI tools?
What aspect of my current matter am I least confident about. And have I researched it myself or relied on AI?
Could I explain the reasoning in the last three judgments I cited without referring to an AI summary?
Am I using AI to do the reading I should be doing myself?

Common questions

Can AI do legal research accurately?

AI tools can surface relevant case law and statutes quickly, but they hallucinate citations, miss jurisdiction-specific nuances, and cannot assess how a specific judge tends to interpret precedent. Every AI-generated legal citation must be manually verified before it appears in any filing, several lawyers have already faced sanctions for failing to do this.

What are the risks of using ChatGPT for legal work?

The most documented risk is citation hallucination, AI tools confidently citing cases that do not exist. Beyond that, AI legal summaries can misstate holdings, overlook recent developments, and strip the contextual caveats that matter in practice. The efficiency gain is real, but the verification burden does not disappear.

Should lawyers use AI for client communication?

AI can draft standard client updates efficiently, but the judgment about what to tell a client. And when, is a core part of legal representation. Clients trust you to filter information based on your reading of their situation. AI-drafted communications approved without that filter risk breaching that trust and, in some jurisdictions, professional obligations.

Will AI replace lawyers?

High-volume, rules-based legal work is already being automated. But litigation strategy, complex negotiation, and advisory work that requires ethical judgment and contextual reasoning are different. The risk is not sudden replacement. It is that lawyers who delegate too much, too early, lose the skills that actually protect them from being replaced.

How can lawyers use AI without compromising their professional judgment?

Treat AI output as a first draft that requires your expertise to evaluate, not a finished product to approve. Use it to cover ground faster, but maintain the habit of reading primary sources directly. The judgment that makes a good lawyer, reading what a case actually turns on, has to remain yours.