Tools
CoCounsel Legal (Thomson Reuters) Review: Westlaw-Grounded Agents
CoCounsel Legal reviewed as of September 2026. The Claude-built rebuild, Brief Builder, Tabular Analysis, Deep Research Verify, $104 to $639 pricing and grounding.
On this page
- From an 18 percent share drop to a Claude-built product in six months
- What the next-generation CoCounsel Legal contains
- The grounding question: what Stanford found, what the sanctions show
- What it costs
- Who uses it
- Who it suits and who it does not
- A verification workflow that uses CoCounsel properly
- Where this leaves you
CoCounsel Legal is Thomson Reuters' AI platform, rebuilt on Anthropic's Claude Agent SDK and Thomson Reuters' own legal model and generally available in the US since 20 August 2026. It is the best fit for US litigators who already pay for Westlaw: Brief Builder, Tabular Analysis and Deep Research Verify are grounded in KeyCite and Westlaw content at published prices of $104 to $639 a user a month. Grounding cuts hallucination; two sanctions involving CoCounsel show it does not end it.
From an 18 percent share drop to a Claude-built product in six months
On 3 February 2026 Anthropic announced a legal plugin for Claude. By the close Thomson Reuters had fallen 18 percent intraday, RELX 14 percent, its steepest one-day decline since 1988, and Wolters Kluwer 13 percent, per Legal.io's account of the fallout. Within three weeks both Harvey and LexisNexis had announced Anthropic integrations, a realignment Legal IT Insider called diverging paths.
Thomson Reuters' path was to build on the thing that had frightened its shareholders. Its 20 August 2026 press release announced the next generation of CoCounsel Legal, built on Anthropic's Claude Agent SDK plus TR's own legal large language model, with Greenberg Traurig as an early deployment. The argument was stated by Thomson Reuters' chief technology officer in Artificial Lawyer's May 2026 coverage: "The control point isn't where work starts. It's whether the output is accurate, grounded in authoritative sources." That sentence is the product.
What the next-generation CoCounsel Legal contains
| Component | What it does | Grounded in |
|---|---|---|
| Westlaw Brief Builder | Drafts brief sections from research, KeyCiting citations as they are inserted | Westlaw, KeyCite |
| Tabular Analysis | Reads up to 10,000 documents against up to 100 questions in one pass | Your documents |
| Deep Research Verify | Runs a research task and checks the authorities it relies on | Westlaw, KeyCite |
| Word drafting agent | Drafts and edits inside Word from Workspace context | Your documents, Practical Law |
| Workspaces | Matter-level context that persists across tasks | Your documents |
| Practical Law | Know-how, standard documents and checklists | Practical Law |
| Evidence integrations | Reveal and Everlaw evidence connected to CoCounsel (Everlaw due autumn 2026) | Review platforms |
Thomson Reuters says 1 million professionals in 107 countries use CoCounsel, the platform has been free to 200-plus law schools since January 2026, and Canada, the UK and Australia get the next generation later in 2026; the UK received the first CoCounsel Legal on 26 January 2026, per TR's January press release. The tax side runs on the same assistant: Checkpoint Edge with CoCounsel has been generally available in the US since July 2024 and TR markets CoCounsel Tax to in-house teams with a vendor claim of 32 percent research time saved.
Two design choices matter for a buyer. Tabular Analysis at 10,000 by 100 is the same shape as Legora's Tabular Review and Anthropic's "every cell cited" diligence grid; it makes CoCounsel a diligence tool, not only a research one. And because CoCounsel is also a partner connector in Anthropic's open-source Claude for Legal plugins, a firm on Claude Team can call Westlaw Deep Research from inside Claude with a CoCounsel subscription; the two products now compete and interoperate at once.
The grounding question: what Stanford found, what the sanctions show
The reason to pay for CoCounsel over a general model is that its answers are built from Westlaw and checked against KeyCite. The evidence on how far that goes is in three places.
Stanford RegLab, 2024. The "Hallucination-Free?" study, later peer-reviewed, tested the previous generation of grounded research tools on 202 expert-scored queries: Westlaw AI-Assisted Research hallucinated 33 percent of the time, Lexis+ AI 17 percent, Ask Practical Law AI about 17 percent, against 43 percent for GPT-4. Fully accurate and grounded answers were 41 percent for Westlaw and 65 percent for Lexis. The products have been rebuilt since, but no vendor has since submitted to a comparable public test; Thomson Reuters, LexisNexis and vLex all declined to take part in Vals.ai's October 2025 legal research study, as LawSites reported.
Vals.ai, February 2025. Where CoCounsel did take part, it scored 89.6 percent on document question-and-answer against a 70.1 percent lawyer baseline and 77.2 percent on summarisation against 50.3 percent, per the Vals report. Lawyers still won on redlining.
The sanctions. In Lacey v. State Farm (C.D. Cal., 6 May 2025) the court imposed $31,100 on K&L Gates and Ellis George after nine of 27 citations proved wrong; the tools used were CoCounsel, Westlaw Precision and Google Gemini, per the AI Incident Database. In Fletcher v. Experian (5th Cir., 18 February 2026) counsel was fined $2,500 for 16 fabricated quotations traced to vLex and CoCounsel, and the court said the problem "shows no sign of abating", per the Vaquill sanctions tracker. The sanctions case list has the full record; the point here is narrow. A grounded tool retrieves real authorities and can still misdescribe them, and a real case cited for a proposition it does not support is the hardest error to see.
Lawyers know this and buy grounding anyway. A LexisNexis survey of 543 legal professionals reported on 1 September 2026 found 81 percent more comfortable with AI grounded in legal sources, up from 70 percent a year earlier. The hallucination guide explains why grounded tools fail differently, not never.
What it costs
| Package (as of September 2026) | Price |
|---|---|
| Published range, CoCounsel Legal | $104–639 per user per month |
| Solo: Westlaw Advantage plus CoCounsel Essentials | $639.20 per month |
| Harvey (estimated, secondary sources) | ~$1,000–1,200 per seat per month, 20-seat minimum |
| Clio Work (vLex content) | $199 per user per month |
| Claude Team | $25 standard / $125 premium per seat per month |
Thomson Reuters reports a Net Promoter Score of 77 for CoCounsel. The published range is the useful number: CoCounsel is the only major legal platform whose entry price a solo can read on a web page, and the pricing guide puts it beside Lexis+ Protégé, whose prices are not published.
Who uses it
ILTA's 2026 survey puts CoCounsel at 44 percent of firms, level with Claude and ahead of Harvey at 43 percent. A Northwestern and NYC Bar survey of 112 federal judges (27 March 2026) found 61.6 percent using at least one AI tool, with Westlaw AI the most common at 38.4 percent, ahead of ChatGPT at 28.6 percent. The Texas Bar's 2026 survey put Westlaw and CoCounsel at 30 percent of AI-using lawyers, against 63 percent for ChatGPT. Thomson Reuters' own Future of Professionals 2026 report (1,816 respondents) found 74 percent of professionals using AI several times a week and 34 percent using unsanctioned tools, which is the gap a firm-licensed, grounded tool is meant to close.
Who it suits and who it does not
Suits: US litigators and litigation teams already paying for Westlaw, for whom Brief Builder and Deep Research Verify are the research workflow they run anyway with the checks moved inside the loop; the brief and motion drafting page shows the pattern. Firms that want a published price and a vendor whose content licence their malpractice carrier has heard of. Diligence teams that want Tabular Analysis without a Harvey commitment.
Does not suit: Transactional teams whose value is in a firm's own precedents rather than Practical Law, where Harvey, Legora or a Claude Project on the DMS does more. Firms outside the US until the next generation reaches Canada, the UK and Australia later in 2026. Lawyers in Austria, Germany or the rest of continental Europe, where Westlaw's domestic coverage is thin and the grounding advantage is only as strong as the content underneath. The direct comparisons are in Harvey vs CoCounsel and CoCounsel vs Lexis+ Protégé, and the rival's own review in Lexis+ with Protégé.
A verification workflow that uses CoCounsel properly
- Frame the question in a Workspace with the matter's documents loaded; write the jurisdiction and the date range into the prompt.
- Run Deep Research Verify, not the plain research skill, so that authorities arrive with their check.
- For every authority the output relies on, open it in Westlaw, read the passage the tool cites, and confirm it says what the draft says; a KeyCite flag tells you treatment, not fit.
- Run KeyCite on the final list yourself; log the date and result.
- Move to Brief Builder only after the authority list is frozen; instruct it to use only the verified list.
- Before filing, run a separate citation check outside CoCounsel, the citation-checking protocol or a dedicated checker, so that the tool that drafted is not the tool that verifies.
- Record the workflow in the file; candour about process is what has separated an unsanctioned filing from a sanctioned one in every 2026 decision.
Jurisdiction: federal, Ninth Circuit; California state law where noted.
Date range: authorities decided on or after 1 January 2018.
Task: identify the controlling standard for [issue], with the three
strongest authorities and the strongest contrary authority.
For each: full citation, KeyCite status, the passage relied on quoted
verbatim, and one sentence on how it applies to the facts in this
Workspace. Do not include any authority you cannot quote.
Where this leaves you
If you are a US litigator on Westlaw, CoCounsel Legal is the natural upgrade and its published price makes the decision easy to model; the risk is not the tool but the habit of trusting a real citation because a grounded system produced it. If you are anywhere else, or your work is transactional, treat it as one option among the platforms rather than the default. Module 2 of the AI Lab for Lawyers course runs the same research question through CoCounsel, Claude and ChatGPT and then verifies all three, which is the fastest way to see what grounding buys and what it does not. The litigation toolkit as a whole, from research to review platforms, is set out in AI for litigators.
Frequently asked questions
How much does CoCounsel Legal cost?
Thomson Reuters publishes a range of $104 to $639 per user per month as of September 2026, depending on the Westlaw and CoCounsel tier bundled. A solo practitioner's Westlaw Advantage plus CoCounsel Essentials package lists at $639.20 per month. Larger firms negotiate. Compared with Harvey's estimated $1,000 to $1,200 per seat, it is inexpensive; compared with Claude Team at $25 to $125, it is the price of the Westlaw content underneath.
Does CoCounsel hallucinate?
Less than an ungrounded chatbot, but yes. Stanford RegLab's 2024 study found Westlaw's AI-Assisted Research hallucinated 33 percent of the time and Lexis+ AI 17 percent, on the previous product generation. In 2025 and 2026 courts sanctioned lawyers in Lacey v. State Farm and Fletcher v. Experian for filings whose tools included CoCounsel. Deep Research Verify and KeyCite reduce the risk; opening every authority yourself removes it.
What is Westlaw Brief Builder?
An agent inside the next-generation CoCounsel Legal that drafts brief sections from Westlaw research, with citations that are KeyCited as they are inserted. It sits beside Tabular Analysis, which reads up to 10,000 documents against 100 questions, Deep Research Verify, matter Workspaces and a Word drafting agent, all generally available in the US since 20 August 2026 and due in Canada, the UK and Australia later in 2026.
Is CoCounsel built on Claude?
Partly. Thomson Reuters rebuilt CoCounsel Legal on Anthropic's Claude Agent SDK, the same framework Anthropic's own legal plugins use, combined with Thomson Reuters' own legal language model and its Westlaw and Practical Law content. CoCounsel is also a partner connector in Anthropic's open-source Claude for Legal, so a firm on Claude can call Westlaw Deep Research through it with a CoCounsel subscription.
Put this into practice
AI Lab for Lawyers: a live, hands-on course for practising lawyers
4 live sessions (8 hours), recordings included, taught by Dr. Niklas Schmidt of Wolf Theiss. Rated 4.7/5 from 381 ratings; alumni from Norton Rose Fulbright, Paul Hastings, Pillsbury and Taylor Wessing.
Related reading
Comparisons
Harvey vs CoCounsel: Platform Play vs Westlaw-Grounded Research
Harvey vs CoCounsel for 2026: Westlaw and KeyCite grounding against firm-wide agents, published prices from $104 versus a 20-seat minimum, and a scenario table.
8 min read
Comparisons
CoCounsel vs Lexis+ Protégé: The Research Incumbents' AI Compared
CoCounsel vs Lexis+ Protégé in 2026: KeyCite against Shepard's, a Claude Agent SDK rebuild against a multi-model harness, prices, jurisdictions, when to switch.
10 min read
Tools
Lexis+ with Protégé Review: Workflows, Vault and the Legal Intelligence Engine
Lexis+ with Protégé replaced Lexis+ AI in February 2026 with 300+ workflows, Shepard's-grounded verification and an agentic engine. What it does, costs and still gets wrong.
8 min read
Use Cases
AI Legal Research in 2026: What Works, What Hallucinates
AI legal research now beats a lawyer baseline on accuracy yet misapplies law in 24% of answers. Which tools are grounded, where errors hide, a routine that holds.
11 min read
Playbooks
How to Verify AI Legal Research: A Citation-Checking Protocol
Verify AI legal research with a five-check protocol: open every authority, match the quote, confirm holding and status, log it. Template, time budgets, tools.
11 min read
Regulation & Ethics
AI Hallucination Sanctions: The Cases Every Lawyer Should Know (2023–2026)
AI hallucination cases reached 2,041 in Charlotin's database by September 2026, 811 involving lawyers. Every sanction from Mata to Couvrette's $110,204, with lessons.
12 min read