CoCounsel Legal

CoCounsel Legal is Thomson Reuters’ current enterprise legal-AI platform for research, drafting, document analysis and multi-step matter work. The current product is built around agentic execution: a lawyer can describe an objective, the system forms an approach, works through research and drafting steps, and returns work product with linked citations. Its strongest structural advantage is access to Thomson Reuters’ own legal corpus: Westlaw primary law, Practical Law guidance and the customer organization’s own knowledge can all inform the workflow. ¹

The product also reaches beyond a standalone chat window. Thomson Reuters documents Microsoft 365 and document-management integrations, matter workspaces that retain context, and an MCP connection that can expose CoCounsel legal research inside Claude under the relevant licensing and permission constraints. Independent reporting supports the broader product direction: Reuters reported one million CoCounsel users in February 2026, while current reporting on Thomson Reuters’ proprietary Thomson model shows the company is adding domain-specific models for workloads such as tabular analysis rather than relying on one general-purpose model for every task. ²³

CoCounsel Legal is an enterprise/contact-sales purchase. The current Thomson Reuters product surface exposes plan and demo paths but not a stable public monetary price, and G2’s current exact-product pricing page likewise describes the offer as contact sales. AiToolMap therefore does not import dollar figures from legacy Casetext pages, reseller estimates or other third-party surfaces. ¹⁴

This opacity is a real buying cost: organizations need a quote to compare total cost against competing legal-AI platforms, and final economics can vary with seats, jurisdiction, Thomson Reuters content subscriptions and negotiated package structure. The upside is that CoCounsel is designed to sit inside a broader professional-content stack rather than operate as an isolated consumer subscription.

The trust architecture is unusually explicit for a legal-AI product. Thomson Reuters says customer content and prompts are not used to train or improve CoCounsel or third-party foundation models, and that partners such as OpenAI and Google are contractually prohibited from using customer data for model training. Current product materials also emphasize source-grounded outputs, visible reasoning steps and human verification rather than autonomous reliance. Integration-specific security documentation adds that usage information may be analyzed for product improvement while user content, prompts and outputs are not used for model training. ⁵⁶

Those are first-party commitments, not a substitute for a customer’s own security and contractual review. Current independent product-level evidence is narrower. G2 reports CoCounsel Legal at 4.8/5 from 75 reviews, the only current exact-product numeric family in the fixed panel that cleanly matches the rebuilt Thomson Reuters surface. Gartner shows 4.0/5 from only two ratings, but its listing remains labeled by Casetext and reflects older product generations, so it is retained as qualitative context rather than used in the rating equation. SoftwareReviews currently says it has insufficient data. A Thomson Reuters-wide Trustpilot profile is also excluded from the numeric equation because it mixes unrelated products, billing and service complaints. ⁴⁷⁸⁹

Historical reporting is useful but must be kept in its lane. VentureBeat and Axios documented earlier CoCounsel model routing and reasoning tests; Ars Technica covered citation-oriented Claude integration while explicitly noting that LLM citation systems still require verification. These sources support the evolution of the architecture, not a claim that the current 2026 product is error-free. ¹⁰¹¹¹²

CoCounsel Legal is best matched to organizations where legal research quality, source traceability, document-heavy workflows and governance matter more than low entry price. It is especially compelling for teams already paying for Westlaw or Practical Law because the proprietary content layer is part of the product’s differentiation rather than an external add-on. Litigation, regulatory analysis, transactional drafting, due diligence and internal knowledge work are all credible target workflows. ¹

It is less attractive for solo users seeking transparent self-serve pricing, teams outside supported Thomson Reuters content jurisdictions, or buyers whose main requirement is a general-purpose assistant rather than a legal work environment. Even in well-grounded legal research, professional review remains mandatory: the system can accelerate retrieval, synthesis and drafting, but responsibility for legal judgment stays with the lawyer.