Westlaw CoCounsel: AI Legal Research Tool Evaluation

What CoCounsel Is — and What It Isn't

Westlaw CoCounsel is Thomson Reuters' AI research assistant, built directly into the Westlaw platform and powered by a combination of GPT-4-class models and Westlaw's proprietary legal content database. It launched in 2023 and has been updated through multiple capability releases since then. The core pitch is straightforward: attorneys can pose research questions in natural language, and CoCounsel returns synthesized answers with citations drawn from Westlaw's corpus.

That framing matters because it defines both the tool's strength and its limits. CoCounsel is not a general-purpose AI assistant — it is a Westlaw-native research accelerator. Users who already subscribe to Westlaw gain access to an AI layer on top of a database they already trust. Users who do not have Westlaw access cannot use CoCounsel independently.

Declared Use Cases

Thomson Reuters positions CoCounsel around five primary workflows. The degree of reliability varies meaningfully across them.

Citation Architecture and Accuracy

CoCounsel's most defensible claim is that its citations link to real Westlaw documents. The system uses retrieval-augmented generation (RAG) against Westlaw's database rather than generating citations from model memory — which is the primary mechanism by which tools like early ChatGPT plugins fabricated case names. When CoCounsel cites a case, that case exists in Westlaw and can be clicked through to verify.

The more nuanced question is whether CoCounsel accurately characterizes what those cases hold. RAG architecture prevents citation fabrication but does not prevent mischaracterization — the model can retrieve a real case and then summarize its holding incorrectly, omit critical limiting language, or fail to surface that the case was subsequently distinguished or overruled. Westlaw's KeyCite integration provides some protection here: CoCounsel is designed to surface negative treatment flags, though practitioners should verify KeyCite status independently rather than relying solely on the AI summary.

Published benchmark comparisons between CoCounsel and competing platforms — including evaluations by law school libraries and legal technology research groups through early 2026 — have generally found CoCounsel competitive on citation accuracy for federal case law research, with performance degrading on state-specific questions in smaller jurisdictions and on regulatory questions where the underlying source material is less consistently structured in Westlaw's database.

Data Privacy and Client Confidentiality

Thomson Reuters has published a customer data protection commitment stating that CoCounsel does not use customer content to train its AI models. Query data and uploaded documents are not retained for model improvement purposes. This is the zero-retention model that most large-firm and in-house legal departments require before deploying any AI tool against client matter files.

The practical implication: attorneys can upload client documents for contract analysis or document review tasks without the same confidentiality risk that applies to consumer-grade AI tools. That said, firms should review Thomson Reuters' current data processing terms and their own engagement letters before uploading sensitive materials — particularly in matters involving trade secrets or regulatory investigations where even transient exposure carries risk.

Pricing Structure

CoCounsel is sold as an add-on to existing Westlaw subscriptions, not as a standalone product. Pricing is structured at the enterprise contract level — Thomson Reuters does not publish per-seat list prices for CoCounsel, and negotiated rates vary significantly by firm size, existing Westlaw relationship, and the specific CoCounsel tier selected.

As of mid-2026, Thomson Reuters offers CoCounsel in at least two tiers: a base tier bundled with Westlaw subscriptions for qualifying plans, and a more capable tier (CoCounsel Core or equivalent branding, which has changed across product cycles) with expanded document upload limits and additional workflow features. Solo practitioners and small firms on entry-level Westlaw plans may find that CoCounsel access is priced out of reach or requires upgrading their base subscription.

Target Audience and Fit

Where CoCounsel Works Well

• Large firms and mid-size firms with existing Westlaw enterprise agreements — CoCounsel is additive to a subscription they already have, and the integration is seamless within the Westlaw interface.
• Associates and senior associates doing initial case law research — the Q&A interface accelerates first-pass research, with citations that can be verified in the same session.
• In-house legal departments that run Westlaw and need to triage large document sets quickly, particularly for contract review and timeline extraction.
• Litigation teams doing deposition preparation on document-heavy matters, where the tool's ability to extract factual sequences from uploaded files saves meaningful time.

Where CoCounsel Is a Poor Fit

• Solo practitioners and small firms on limited Westlaw plans — the cost structure and bundling model makes CoCounsel harder to access without upgrading a base subscription that may already be expensive relative to firm revenue.
• Practitioners who primarily use Lexis+ or Bloomberg Law — there is no standalone CoCounsel product; it lives inside Westlaw and cannot be purchased independently.
• Transactional practices focused on drafting new documents rather than researching existing law — CoCounsel's drafting capabilities are limited compared to purpose-built contract drafting tools like Spellbook.
• Matters requiring deep regulatory research in non-Westlaw-indexed sources — CoCounsel's knowledge is bounded by Westlaw's database, which does not uniformly cover all agency guidance, foreign law, or specialized regulatory corpora.

Known Accuracy Limitations

Several limitations have been documented through independent testing and practitioner reports through mid-2026:

Professional Responsibility Considerations

Using CoCounsel does not transfer professional responsibility. The supervising attorney remains responsible for the accuracy of research cited in filings, opinions, and client advice. Several state bar guidance documents issued through 2025 and 2026 have addressed AI-assisted research specifically, generally requiring that attorneys verify AI-generated citations against primary sources before relying on them.

The competence obligation under Model Rule 1.1 — and its state equivalents — applies to the attorney's understanding of the tool's limitations, not just the output it produces. Attorneys using CoCounsel should be able to explain how the tool works at a basic level, understand the RAG architecture's limitations, and have a defined verification workflow before outputs are used in client matters.

A growing number of courts have issued standing orders requiring disclosure when AI tools are used in preparing filings. Practitioners should check whether their jurisdiction has such requirements before submitting CoCounsel-assisted work product.

Competitive Position

CoCounsel's primary direct competitors in AI-assisted legal research are Lexis+ AI (LexisNexis) and Bloomberg Law's AI features. The competitive dynamic largely maps to existing database relationships: firms that have built workflows around Westlaw will find CoCounsel the natural AI layer; firms on Lexis will evaluate Lexis+ AI first.

Where CoCounsel has a structural advantage is database depth — Westlaw's case law corpus is broadly regarded as comprehensive for US federal and state courts. Where it faces competitive pressure is in document-centric workflows: Harvey, which operates independently of any single database, has made inroads at large firms for tasks like contract review and due diligence where the research database matters less than the document analysis capability.

Evaluation Summary