The Legal Research Platform Wars: How AI Is Changing the Way Lawyers Find and Use Case Law
Legal research has always been expensive, slow, and unforgiving. Miss a controlling case in the Ninth Circuit and you've committed malpractice. Spend six hours chasing a circuit split that a paralegal already mapped last week and you've blown the client's budget. AI-assisted research promised to...
By Andy Armstrong | The Legal Stack | April 24, 2026
Legal research has always been expensive, slow, and unforgiving. Miss a controlling case in the Ninth Circuit and you've committed malpractice. Spend six hours chasing a circuit split that a paralegal already mapped last week and you've blown the client's budget. AI-assisted research promised to fix both problems simultaneously. Two years into the serious commercial deployment of these tools, the honest verdict is more complicated — and the platform you choose matters enormously.
Here's where things actually stand.
Westlaw Precision: The Incumbent Plays Defense
Thomson Reuters spent serious money building Westlaw Precision, and the KeyCite integration remains its most defensible moat. When you ask Precision a research question, the citations it returns carry real-time validity flags. The system doesn't just find Chevron U.S.A., Inc. v. Natural Resources Defense Council — it tells you immediately that Loper Bright Enterprises v. Raimondo (2024) overruled it. That doctrinal awareness, baked into the citation layer, is genuinely hard to replicate.
The problem is cost. Westlaw Precision runs between $500 and $1,200 per user per month depending on firm size and practice module. For a solo practitioner or a boutique firm, that's a meaningful line item every billing cycle. The AI summarization features — particularly the document upload and memo drafting tools added in early 2025 — are competent but rarely surprising. Precision excels at confirming what you already suspect. It is less impressive at finding what you didn't know to look for.
Hallucination risk: Low for citation retrieval, moderate for synthesized summaries. The underlying index is real. The prose wrapped around it occasionally overstates.
Lexis+ AI: Better at Analysis, Shakier on Citations
LexisNexis has been more aggressive about integrating large language model capabilities directly into the research workflow. The Lexis+ AI "Ask" interface genuinely converses — it handles follow-up questions, remembers context within a session, and produces memo-style outputs that solo practitioners can turn into client deliverables faster than anything Westlaw currently offers.
The tradeoff is citation reliability. In testing conducted across several mid-size litigation firms during Q4 2025, Lexis+ AI produced confidently-cited but slightly off-point cases at a higher rate than Westlaw Precision — not fabrications exactly, but citation drift, where the case exists and the principle is roughly right, but the quoted language doesn't appear at the pinpoint cite provided. For transactional work, this is annoying. For a brief, it's catastrophic. Always shepardize independently before filing.
Cost: Comparable to Westlaw, with more aggressive discounting for small firms and solo practitioners who negotiate.
Casetext (Now Under Thomson Reuters): The Acquisition Changed the Calculus
Casetext's CARA AI product was genuinely innovative — upload a brief, get a list of cases the opposing party cited that you haven't addressed. Thomson Reuters acquired Casetext in 2023 for $650 million, and the integration into the Westlaw ecosystem has been ongoing. The standalone Casetext product still exists as of this writing but the roadmap points toward consolidation.
If you were betting on Casetext as an independent disruptor, that bet is effectively over. The technology lives on inside Westlaw Precision's document analysis features. The lesson here is that in legal tech, "acquired by Thomson Reuters" is often where startup differentiation goes to retire gracefully.
Harvey: The BigLaw Tool Trying to Go Downstream
Harvey AI has positioned itself as infrastructure rather than a research tool specifically, but its legal research module — particularly in partnership with Allen & Overy (now A&O Shearman) and other early BigLaw adopters — has demonstrated something the incumbent platforms struggle with: multi-jurisdictional synthesis at scale.
Ask Harvey to map how courts in Delaware, New York, and California have treated material adverse change clauses post-AB Stable VIII LLC v. MAPS Hotels and Resorts (2020), and you get a structured comparative analysis that would take a junior associate a full day. The accuracy on established doctrine is strong. The risk is at the edges — novel questions where the model extrapolates from analogy rather than precedent. Harvey is transparently a tool that requires a senior lawyer to supervise the output.
Pricing has historically been enterprise-only, but 2025 saw mid-market packages emerge. Expect more of this.
The Emerging Challengers Worth Watching
Two platforms deserve attention before they're either acquired or ignored:
Spellbook (transactional focus, contract drafting with research integration) has quietly built a loyal user base among M&A associates who want clause-level research contextualized within a live document.
EvenUp — primarily a plaintiff's PI tool — demonstrates something important: vertical specialization beats general capability for high-volume practice types. A general-purpose AI research tool will never understand PI case value the way a tool trained specifically on settlement databases will.
The Honest Cost-Per-Query Breakdown
| Platform | Approx. Monthly Cost (Solo) | Hallucination Risk | Best For |
|---|---|---|---|
| Westlaw Precision | $500–$800 | Low | Litigation, citation-critical work |
| Lexis+ AI | $450–$750 | Moderate | Analysis, drafting, small firms |
| Harvey | $1,000+ (enterprise) | Low-Moderate | BigLaw, complex multi-jurisdiction |
| Casetext (standalone) | $150–$300 | Moderate | Budget-constrained practitioners |
What Practitioners Should Actually Do
Stop treating this as a binary switch decision. The lawyers winning at legal research in 2026 are using Westlaw Precision for citation verification and validity checking, and a cheaper AI-native tool — Lexis+ AI or Casetext — for initial issue spotting and synthesis. The two-tool workflow is messier than a single vendor relationship but meaningfully better on accuracy-per-dollar.
The deeper issue is professional responsibility. Model Rules 1.1 (competence) and 5.3 (supervision of nonlawyers) increasingly get cited by bar ethics committees in the same breath as AI research outputs. The California State Bar's 2025 formal opinion on AI use in practice made this explicit: you own every citation in every document you file, regardless of which platform generated it.
The platforms are improving fast. The professional obligations aren't changing at all.
Andy Armstrong covers legal technology and practice innovation for The Legal Stack. He has no financial relationship with any platform mentioned in this article.