Casetext (CoCounsel) vs LexisNexis Protégé

Casetext (now Thomson Reuters CoCounsel) versus LexisNexis Protégé is the proxy fight between TR and LexisNexis for the AI-on-legal-research market. Both ship AI agents grounded in their respective databases — Westlaw for CoCounsel, Lexis Advance for Protégé. The decision is rarely about the AI itself; it’s about which research stack your firm already lives in.

Where Casetext wins

• Earlier AI maturity. CoCounsel shipped real GPT-4-grade workflows in 2023, before TR acquired Casetext. Protégé caught up later, and the lineage shows in product polish around skills like Document Review and Deposition Prep.
• Westlaw and KeyCite linkage. If your firm’s research workflow is Westlaw-native, every CoCounsel output drops cleanly into the same database, with KeyCite treatment and a unified citation trail.
• Skill-based UX. CoCounsel’s “skills” model (Review Documents, Legal Research Memo, Prepare for Deposition) is opinionated and easy to adopt. Protégé’s surface is broader but less prescriptive.

Where Protégé wins

• Lexis ecosystem fit. For Lexis-centric firms, Protégé inherits Shepard’s signals, secondary sources (Matthew Bender, Mealey’s), and existing SSO. Switching to CoCounsel for Lexis shops means changing research stacks too.
• Agentic depth. Protégé has invested heavily in multi-step agents — research plus drafting plus citation-checking in one pass. CoCounsel’s skills are more discrete.
• International coverage. Lexis has stronger non-US footprints in several jurisdictions. For UK, Canadian, and certain APAC firms, Protégé reflects local content better.

Pricing reality

Both bundle into the parent’s enterprise contracts. If you’re a Westlaw / Practical Law shop, CoCounsel’s marginal cost is reasonable. If you’re a Lexis shop, Protégé’s is. Cross-buying is rare and rarely worth it — you end up paying twice for the underlying research database, which is the actual moat.

Verdict

• Pick Casetext if your firm runs on Westlaw, your work is litigation- and research-heavy, and you want a polished skill-based AI surface.
• Pick Protégé if your firm runs on Lexis, you value Shepard’s, and you want deeper agentic workflows tied to the Lexis content stack.
• Don’t run both. This is the rare market where the AI is downstream of the research database. Pick the database, take the AI that comes with it.

The single mistake to avoid: switching research providers for the AI. The content layer is the moat; the model layer will commoditize.