The Three-Question Test Your Lawyer Should Run on Any AI Output
Why Lawyers Can’t Just Trust AI at Face Value
Artificial intelligence tools are becoming a regular part of legal work. They help with research, drafting documents, summarizing case files, and pulling together information quickly. But speed does not always mean accuracy. And in legal practice, inaccurate information can cause serious harm to clients, damage a lawyer’s reputation, and even lead to professional discipline.
The core problem is simple: AI systems can sound very confident while being completely wrong. They generate text that looks polished and professional, even when the underlying facts, citations, or legal reasoning are off base. This is why attorney due diligence is not optional when AI is involved. It is a professional requirement.
The good news is that you do not need a complicated system to catch most AI errors. A straightforward three-question test can help lawyers evaluate any AI output before relying on it. This kind of structured legal verification process protects clients and keeps legal professionals on solid ground.
The Growing Role of AI in Legal Work
Before getting into the test itself, it helps to understand what is actually at stake. AI tools used in legal settings can make several types of errors:
- Hallucinated citations: AI can generate case names, docket numbers, and court decisions that do not exist.
- Outdated legal information: Laws change, and AI training data has a cutoff date that may mean it misses recent developments.
- Misapplied legal standards: AI may apply the law from one jurisdiction incorrectly to a case in a different state or country.
- Incomplete analysis: AI might summarize only part of a legal issue, leaving out critical nuances or exceptions.
- Factual inaccuracies: Basic facts about statutes, regulations, or legal processes can be wrong.
Each of these mistakes, if left unchecked, can become a serious problem in court filings, client advice, or contract work. The three-question test is designed to catch these issues before they cause real damage.
Question One: Can I Find the Source?
The first question every lawyer should ask about any AI output is whether the information can be traced back to a real, verifiable source. This is the foundation of legal verification.
AI tools often present information as if they have read and understood specific documents, cases, or statutes. But that does not mean the source actually exists or says what the AI claims it says. Lawyers have already faced embarrassing situations in court where they submitted briefs citing cases that were entirely fabricated by an AI tool.
How to Apply This Question
When reviewing AI output, pull every citation and look it up independently. Use trusted legal research platforms like Westlaw, LexisNexis, or official government databases. Check that:
- The case or statute actually exists with the name and number the AI provided.
- The quote or holding the AI described matches what the actual document says.
- The source is from the correct jurisdiction and court level.
- Any regulatory reference points to a currently active rule, not one that has been repealed or amended.
If the AI output does not cite sources at all, that is a signal to be extra careful. Unsourced legal claims require full independent research before they can be used. The AI quality of the output, no matter how well-written it appears, means nothing if the underlying information cannot be confirmed.
Question Two: Is This Still the Law?
The second question focuses on currency. Legal rules change. Courts overturn precedents. Legislatures amend statutes. Agencies update regulations. An AI tool trained on data from even a year ago may be giving you outdated information without any indication that things have changed.
This is not a flaw that better AI writing will fix. It is a structural limitation of how these tools are built. They work from training data with a specific cutoff, and the law keeps moving after that point.
How to Apply This Question
After confirming that a source exists, verify that it is still current. Specifically:
- Run a Shepard’s or KeyCite check on any case to confirm it has not been overturned, distinguished, or criticized in ways that affect how it applies.
- Check the current version of any statute on the official legislative website to confirm no amendments have changed the relevant language.
- Look up any regulations on official agency websites or the Code of Federal Regulations to see if there have been recent updates.
- Pay particular attention to areas of law that change frequently, such as tax law, immigration rules, environmental regulations, and employment standards.
Attorney due diligence on currency is especially important when the AI output sounds very settled and definitive. Sometimes the most confident-sounding statements are the ones that reflect law that has since shifted. A factual and careful review process will catch this before it becomes a problem in a client matter.
Question Three: Does This Apply to My Client’s Situation?
The third question is about relevance and fit. Even if AI output points to a real source that is currently good law, that does not automatically mean it applies to the specific facts and jurisdiction of your client’s case.
AI tools are generalists. They pull together broad information about legal topics, but they do not know your client’s specific facts unless you have provided them in detail. And even then, they may apply general principles without accounting for the specific wrinkles that make your case different.
How to Apply This Question
Work through a careful analysis that compares the AI output to your actual case. Consider:
- Jurisdiction: Is the case law from the right state or federal circuit? Does the statute apply in your client’s location?
- Facts: Does the legal standard the AI described actually fit the facts of your matter, or is the AI drawing an analogy that does not hold up under closer review?
- Client-specific context: Are there aspects of your client’s situation — their business structure, prior legal history, contract terms, or industry — that change how the law applies?
- Procedural posture: Is the AI discussing the law in the right procedural context? A rule that applies at trial might work differently at the summary judgment stage or on appeal.
This third question often requires the most actual lawyering. It is where professional judgment, experience, and client knowledge matter most. AI quality can be high in terms of how something is presented while still being a poor fit for the specific task at hand.
How to Build This Into Your Practice
Running this three-question test does not have to be complicated or time-consuming. The key is to make it a consistent habit rather than something you do only when you feel uncertain. Here are some practical ways to build it into your workflow:
- Create a simple checklist: Put the three questions on a one-page document that your team uses whenever AI output is incorporated into a work product.
- Document your verification steps: Keep brief notes on what you checked and where, especially for matters where AI played a significant role.
- Train staff and associates: Anyone using AI tools in your office should understand these questions and how to apply them.
- Do not skip the test under time pressure: Deadlines create pressure to cut corners, but the consequences of unverified AI output in a filed document or client letter are far worse than the time spent on verification.
- Revisit AI output if facts change: If the facts of a matter develop after you have used AI to research an issue, run the test again with the new information in mind.
What This Test Does Not Do
It is worth being clear about the limits of this approach. The three-question test is a practical tool for catching the most common and serious AI errors in legal work. It is not a complete substitute for thorough legal research and professional analysis.
There are complex matters where AI output can pass this basic test and still be inadequate for the depth of analysis a client’s situation requires. Creative legal arguments, novel fact patterns, and high-stakes litigation all demand more than verification. They require the kind of careful, experienced thinking that no AI tool currently provides.
The test is a floor, not a ceiling. It sets a minimum standard for attorney due diligence when AI is involved. Good legal work will often go well beyond it.
The Professional Responsibility Angle
Lawyers have ethical obligations that go beyond simply trying to get the right answer. Rules of professional conduct in most jurisdictions require competence, diligence, and candor to the court. Using AI output without verification can run into all three of these obligations at once.
Bar associations and courts are increasingly aware of AI-related errors in legal work. Some courts have already adopted specific rules requiring disclosure when AI has been used to prepare filings. More are expected to follow. Running a consistent legal verification process is not just good practice — it is increasingly the professional standard.
Lawyers who treat AI as a starting point and not a final answer, and who apply structured review habits like the three-question test, are in a much stronger position both professionally and practically.
Final Thoughts
AI tools can genuinely help lawyers work more efficiently. They are not going away, and there is no reason to avoid them entirely. But the responsibility for the accuracy and quality of legal work stays with the lawyer, not the machine.
The three-question test — can I find the source, is this still the law, and does this apply to my client — gives any attorney a simple, reliable framework for maintaining quality control over AI output. It protects clients, guards against professional liability, and keeps the practice of law grounded in the kind of careful, factual analysis that legal work has always required.
Using AI well means using it with eyes open. These three questions are a good way to keep them that way.
