Legal research is where AI's strengths and weaknesses collide most dramatically. AI can synthesize complex legal concepts, identify relevant frameworks, and generate structured research memos in minutes. It also regularly fabricates case citations, invents holdings, and presents fiction as authority.
Using AI for legal research requires a specific discipline: extract its analytical capabilities while rigorously verifying everything it presents as fact.
Disclaimer: AI-generated legal research must be independently verified. AI can and does generate fabricated citations, misstate holdings, and present outdated law as current. No AI output should appear in any filing or work product without thorough verification by a qualified attorney.
The Hallucination Problem
Before the prompts, the most important section in this guide.
AI does not look up cases in a database. It predicts text based on patterns in training data. When asked for citations, it generates text that looks like a legal citation — but may combine a real case name with a wrong reporter, invent a holding from a real case, or fabricate an authority entirely.
Real consequences: Multiple attorneys have faced court sanctions — fines, mandatory education, public reprimands — for submitting briefs with AI-fabricated citations.
The Verification Protocol
Every piece of AI legal research must pass these checks:
- Existence: Look up every case in Westlaw or LexisNexis
- Citation accuracy: Verify reporter, volume, and page number
- Holding verification: Read the case and confirm the AI's description
- Current status: Run KeyCite or Shepard's — is it still good law?
- Relevance: Does the case actually support the stated proposition?
- Statutory currency: Confirm statutes have not been amended or repealed
Case Law Research Prompts
Legal Framework Mapping
Start with the conceptual landscape, not specific cases.
You are a legal research attorney.
Research: [LEGAL QUESTION]
Jurisdiction: [STATE or FEDERAL CIRCUIT]
Map the framework WITHOUT specific case citations:
1. What area(s) of law does this implicate?
2. Key legal doctrines or tests
3. Relevant statutes or regulations
4. Competing policy considerations
5. How the landscape has evolved recently
6. Circuit splits or jurisdictional differences
Do NOT cite cases yet — I will verify separately.This prompt deliberately avoids citations in the first pass. The conceptual framework is where AI adds the most value with the least risk.
Research Path Generation
Suggest 4-6 specific research paths for [LEGAL QUESTION]:
For each:
1. The legal theory or argument
2. Type of authority to find (cases, statutes, regulations,
secondary sources)
3. Suggested Westlaw/LexisNexis search terms
4. Key concepts or tests to understand
5. Likely counterarguments to research
Mark any specific cases or statutes with [VERIFY].Case Law Discussion
Discuss key judicial decisions shaping [LEGAL ISSUE] in
[JURISDICTION].
For each case: name and approximate year [VERIFY], court,
key relevant facts, holding and reasoning, how it advanced
or limited the law, current significance.
Then: What patterns emerge? Where is the law settled vs.
unsettled? What fact distinctions determine outcomes? What
arguments do courts find most persuasive?
CRITICAL: Flag uncertainty clearly. It is better to say
"there was a case around [YEAR] involving [FACTS] — verify"
than to present a fabricated citation confidently.Adverse Authority Identification
I am arguing that [POSITION].
Identify the strongest potential adverse authority:
1. Legal theories against this position
2. Fact patterns where courts ruled against similar arguments
3. Common judicial criticisms
4. Doctrines or defenses that undermine the position
For each: describe what to look for, suggest search
strategies, and explain how to distinguish or address it.
Mark any cases with [VERIFY].Research Memo Draft
Draft a research memo on [QUESTION] for [JURISDICTION].
Structure: Question Presented, Short Answer, Statement of
Facts, Discussion (IRAC for each sub-issue), Open Questions,
Conclusion.
Use signal phrases correctly (see, see also, cf., but see).
Mark ALL citations [VERIFY]. Prioritize accurate legal
analysis over citation precision.Statutory Analysis Prompts
Statutory Interpretation
Analyze [STATUTE — full citation] as applied to [SCENARIO].
1. TEXT: plain reading, key definitions, ambiguities
2. STRUCTURE: fit within broader scheme, related provisions,
safe harbors
3. PURPOSE: legislative intent, regulatory objectives
4. INTERPRETATION: agency guidance, judicial interpretation
in [JURISDICTION] [VERIFY all]
5. APPLICATION: elements met, elements contested, risk
6. PRACTICAL: compliance steps, common failures, record-
keeping requirements
Flag the version analyzed — I will confirm currency.Multi-Statute Interaction
Analyze how these statutes interact for [SCENARIO]:
- Statute 1: [CITATION]
- Statute 2: [CITATION]
Do provisions conflict? Which controls? Preemption issues?
Cumulative requirements? Could compliance with one cause
non-compliance with another? What is the most conservative
approach satisfying all?
Note genuinely unsettled areas of interaction.Legislative History Research
Research legislative context for [STATUTE] regarding
[SPECIFIC AMBIGUITY].
Cover: the problem the legislation addressed, legislative
findings, committee report language, rejected alternatives,
subsequent amendments, and agency statements.
Note: how much weight does [JURISDICTION] give legislative
history? Is the textualist vs. purposivist distinction
relevant here?
Mark all references with [VERIFY].Regulatory Research Prompts
Regulatory Framework Mapping
Map the complete regulatory framework for [BUSINESS ACTIVITY]
in [JURISDICTION].
FEDERAL: agency, applicable CFR sections, licensing, ongoing
obligations, enforcement focus.
STATE (for [STATES]): state agency, additional requirements,
licensing.
INDUSTRY-SPECIFIC: SROs, standards, professional licensing.
For each: mandatory vs. advisory, deadlines, penalties, how
compliance is demonstrated.
Mark all citations [VERIFY] — regulations change frequently.Enforcement Trend Analysis
Analyze enforcement trends for [REGULATORY AREA] over the
past 2-3 years:
- Direction (more aggressive, stable, de-emphasizing?)
- Common violation types targeted
- Typical actions and penalty ranges
- Industries most frequently targeted
- Published enforcement priorities
- Emerging focus areas
For [OUR COMPANY TYPE]: highest-priority compliance areas,
proactive risk-reduction steps, safe harbor programs.
Flag any specific actions or data I should verify.Agency Guidance Interpretation
Analyze [AGENCY] guidance on [TOPIC]:
1. Formal rules — current, pending, recent changes
2. Informal guidance — FAQs, policy statements, advisory
letters, leadership statements
3. Enforcement as guidance — consent orders, enforcement
actions signaling priorities
4. Practical interpretation — what the agency expects,
gray areas, conservative vs. aggressive approaches
5. Future direction — pending rulemaking, enforcement
priority signals
Mark all specific documents and actions with [VERIFY].Research Organization Prompts
Multi-Issue Research Summary
Synthesize research across these issues in [MATTER]:
1. [Issue 1] 2. [Issue 2] 3. [Issue 3]
For each: current standard in [JURISDICTION], position
strength (Strong/Moderate/Weak/Uncertain), key supporting
authority [VERIFY], key adverse authority [VERIFY], strategic
recommendation, further research needed.
Overall: how issues interact, whether strength in one
compensates for weakness in another, priority for further
research.Research Gap Identification
Review our research on [ISSUE] and identify gaps:
1. Unexplored legal arguments
2. Unaccounted factual scenarios
3. Helpful secondary sources (treatises, law reviews)
4. Overlooked admin decisions or agency interpretations
5. Adequacy of adverse authority research
6. Jurisdictional issues overlooked
7. Pending legislation that could change the analysis
For each: why it matters, steps to fill it, importance
(critical / helpful / completeness only), and search
strategies.Where AI Adds the Most Research Value
Not all legal research tasks benefit equally from AI. Understanding where AI helps most guides how to allocate your time.
High-value AI tasks:
- Mapping an unfamiliar area of law before diving into databases
- Generating search term variations you would not have considered
- Organizing multi-issue research into a coherent memo structure
- Identifying potential arguments from both sides of a dispute
- Summarizing long regulatory guidance into actionable summaries
- Drafting research outlines that an associate can then flesh out
Low-value or risky AI tasks:
- Finding the controlling case in a specific jurisdiction (use Westlaw/Lexis)
- Confirming whether a case is still good law (use KeyCite/Shepard's)
- Identifying every case in a line of precedent (use citator tools)
- Producing final citation-ready research memos without verification
- Analyzing very recent legal developments (training data cutoffs)
The pattern: AI excels at the conceptual and organizational layers of research. Traditional databases excel at the citation-level accuracy and currency layers. The best research combines both.
The Practical Verification Workflow
Step 1: Flag Everything
Mark every factual claim: case names, citations, statutory references, legal standards, legislative history claims, agency positions.
Step 2: Verify Primary Sources
Start with the most heavily relied-upon authorities. Look up in Westlaw or LexisNexis: Does it exist? Is the citation right? Does it say what AI claims? Is it still good law?
Step 3: Fill Gaps
AI research always has gaps. Use gap identification prompts, then fill through traditional research.
Step 4: Update the Analysis
Remove fabrications, correct misstated holdings, add missed authority, update the analysis to reflect verified law.
Step 5: Final Quality Check
Every citation verified, analysis reflects cited authority, no AI fabrications remain, research is current.
Building Verification Into Your Practice
The verification workflow is faster than it sounds. Most experienced attorneys can check a case citation in under a minute using Westlaw or LexisNexis. A research memo with 15-20 citations takes 20-30 minutes to verify — far less than the hours it would take to produce the same conceptual analysis from scratch.
The key is building verification into the workflow from the start, not treating it as an optional final step. When verification is standard procedure, AI legal research becomes a genuine time-saver. When it is skipped, it becomes a risk.
Making AI Legal Research Work
AI legal research is not a shortcut — it is a different workflow that trades citation-level accuracy for speed in the conceptual and organizational layers.
Used correctly: AI identifies frameworks you might not have considered, generates search strategies, organizes multi-issue research, drafts memos for verification, and spots research gaps.
Used incorrectly: AI introduces fabricated authority into your work product, leading to sanctions, malpractice exposure, and credibility loss.
The difference is verification discipline. Build it into your workflow from the start, and AI becomes a powerful accelerator. Skip it, and AI becomes a liability.
For prompt templates across other legal tasks, see AI prompts for lawyers, AI contract analysis, and AI for compliance. Build custom legal research prompts with SurePrompts' prompt generator and save them in the Template Builder.
FAQ
How do I cite AI-assisted research in legal documents?
You do not cite AI as a source. AI is a tool, not an authority. After using AI to identify frameworks and hypotheses, verify through authoritative databases and cite primary/secondary sources directly. Some jurisdictions require disclosure that AI was used in preparing filings — check local rules and court orders for AI disclosure requirements.
Is AI better for some types of legal research than others?
AI excels at conceptual research: identifying frameworks, explaining how doctrines relate, generating strategies, and organizing multi-issue analysis. It is weakest at citation accuracy and current regulatory status. For conceptual tasks (understanding unfamiliar law, mapping arguments for novel issues), AI adds significant value. For citation-dependent tasks (finalizing briefs, preparing filings), traditional databases remain essential.
How do I handle confidentiality when researching client matters?
Never enter actual client names, case numbers, or privileged communications. Use anonymized hypotheticals — replace names with placeholders, describe facts generically, omit identifying details. For sensitive matters, consider whether the legal question itself could identify the client. Use enterprise AI platforms with DPAs that prohibit training on inputs.