Module 2.1 · Topic 5
Prompting for Legal Analysis and Research
Bottom Line Up Front: Legal analysis requires prompting precision, citation verification, and document review discipline as professional obligations. All foundational techniques plus legal-specific disciplines apply.…
5.1 Structuring Prompts for Legal Research Tasks
Legal research prompts should specify relevant jurisdiction, statutes, and courts rather than relying on the AI to make those calls. Generic research approaches will generally fail to deliver what your need.
Why it works:
Law is jurisdiction-specific. Specifying jurisdiction eliminates risk of mixed jurisdictions. Prompt precision determines research accuracy.
Sample prompt:
What to expect:
Jurisdiction-specific precision with verifiable citations.
5.2 Citation Verification and Source Validation Prompts
Never rely on AI citations without verification. Ask AI (in the prompt, not after) to identify sources and distinguish holdings from dicta, then still approach AI-generated citations with caution.
Why it works:
AI citations can be fabricated. Verification checkpoints are essential.
Verification Steps:
- Request verification: "For each citation, verify it exists and provide a link to the source you used."
- Holdings vs. dicta vs. commentary: "I want holdings from the judgment text, not just summaries or commentary, and don't treat dicta as holding."
- Confidence levels: "Identify all sources, and note your confidence for each source. Saying 'uncertain' is better than a confident guess."
- Independent verification: Unless you are using an AI explicitly described as drawing from a legal database (Westlaw, Google Scholar, etc...) or court website, it's your job to confirm every citation. Even then, read the source to confirm the case says what the AI claims it said.
Critical Warning: Verification is Non-Negotiable
AI language models have hallucinated completely fabricated cases and citations. Mata v. Avianca Inc. (S.D.N.Y. 2023) is a well-known example where an attorney cited nonexistent cases generated by AI, resulting in sanctions. The Eleventh Judicial Circuit Administrative Order No. 26-04 (2026) now requires judges to treat fabricated citations as potential grounds for discipline. Before filing any document citing AI-generated legal research, verify every citation in independent sources. This is not optional—it is a professional obligation. The Stanford Legal RAG Hallucinations Study found hallucination rates of 1 in 6 or higher in legal AI systems, making verification essential.
5.3 Jurisdiction-Specific Prompting Strategies
Specify jurisdiction and statutes so AI applies the right legal framework.
Why it works:
Legal rules vary by jurisdiction. Specifying jurisdiction eliminates ambiguity. Critical for state-variable practice areas.
Sample prompt:
What to expect:
Analysis more likely to be grounded in your jurisdiction's law and precedent.
5.4 Prompting for Document Review and Analysis
For document review, define "red flags" precisely and ask AI to explain reasoning for consistency.
Why it works:
Defining red flags upfront creates verifiable frameworks. Critical for contracts where missed clauses matter.
Sample use cases with AI model-type tags:
- Contract Red-Flag Analysis Thinking/Reasoner: Define red flags, upload contract, ask AI to identify and explain each, suggest fixes.
- Compliance Review Deep Research: Define standards (e.g., GDPR), upload policy, identify gaps and remediation.
- Agreement Consistency Fast/Chat: Upload multiple agreements, identify inconsistencies across non-competes, confidentiality, IP.
- Discovery Prioritization Thinking/Reasoner: Upload sample documents, categorize by case relevance, flag privilege issues.
Professional Responsibility:
AI legal analysis requires client disclosure and ethics compliance. Check bar association guidance. See Module 7.1 for liability and ethics context.