Westlaw Precision AI vs Lexis+ AI: Research Compared
Start with a naming correction, because it changes the comparison. "Westlaw Precision AI" and "Lexis+ AI" are both legacy names in 2026. Thomson Reuters has moved its research line to Westlaw Advantage, with the AI work bundled through CoCounsel, and LexisNexis renamed Lexis+ AI to Lexis+ with Protege in February 2026. On the current products, the substantive differences are these: Westlaw grounds its answers in West's editorial layer (KeyCite, headnotes, the Key Number System) and sells the AI as part of a research or CoCounsel bundle, while Lexis grounds its answers in Shepard's citation validation plus Practical Guidance and now adds a Shepard's Verify pass over AI output. Thomson Reuters publishes list pricing for small firms, LexisNexis does not publish any. And the accuracy percentages you see quoted across comparison posts almost all trace back to one 2024 study that tested neither of the products either company sells today.
We do not sell either platform, and no vendor pays to influence what appears here. That is the whole reason this page can lead with the naming problem instead of burying it.
What does each platform's AI actually do?
On the Thomson Reuters side, the current research platform is Westlaw Advantage. Its published feature set includes Deep Research, the Litigation Document Analyzer, Complaint Analyzer, Claims Explorer, and AI Jurisdictional Surveys, sitting on the classic Westlaw furniture: KeyCite, the West Key Number System, and headnotes. Alongside it, Thomson Reuters sells CoCounsel Essentials for document analysis and drafting, and CoCounsel Legal as the combined stack. The lineage is easy to lose track of, and a Kathrine R. Everett Law Library post at UNC lays it out cleanly in a history of Westlaw's rebrands from WestlawNext through Precision to Advantage: Precision arrived in 2022 with new filters, AI-Assisted Research followed in 2023, CoCounsel arrived as a workflow product, and CoCounsel Legal launched in August 2025. So when a comparison post reviews "Westlaw Precision AI," it is reviewing a 2022 interface layered with a 2023 research feature.
On the LexisNexis side, Lexis+ AI is now Lexis+ with Protege. Protege is the assistant, and the platform around it covers research, full document drafting for both transactional and litigation work, review tables for large document sets, shared Workrooms, and a secure Vault for storing and analyzing document sets. It connects to iManage, SharePoint and NetDocuments so it can draft from a firm's own knowledge base, and it exposes multiple underlying model choices rather than one fixed model.
Functionally, the two have converged more than either sales team will admit. Both now run research, drafting, document analysis, and citation checking in one workspace. Both index a deep proprietary corpus of primary law and secondary commentary. If your evaluation is a feature bake-off, expect a near tie and a lot of wasted pilot time.
How do their grounding approaches differ?
This is the difference that survives model releases, so it is where an evaluation should actually start.
Westlaw's grounding is editorial. West has spent more than a century classifying American case law by legal issue, and headnotes, the Key Number System and KeyCite are the artifacts of that work. When Westlaw's AI retrieves, it retrieves against a corpus that human editors have already labeled and cross-referenced. The strength is structural: the system has an opinionated map of what a case stands for, not just its text. The weakness is that the map is West's, and their taxonomy becomes your research vocabulary.
Lexis's grounding leans on Shepard's, which is a treatment-history apparatus rather than an issue taxonomy. It tells you whether an authority is still good and how later courts have treated it. In the current release LexisNexis has pushed that further with Shepard's Verify, described as a check that runs across citations in AI responses, documents and drafts, flagging statements that lack support. That is a meaningfully different design choice: rather than only grounding generation in trusted content, it adds a verification pass over the generated output.
Neither approach is obviously better. Editorial classification helps most when you are trying to find the right line of authority. Treatment validation helps most when you already have authority and need to know whether it holds. Most litigators do both jobs weekly, which is why practitioners who have used both platforms tend to describe differences in feel rather than in capability.
What do the accuracy studies really show?
Here is where most comparison content goes wrong. Percentages get quoted as if they describe today's products. They do not.
The one rigorous public evaluation is Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools from Stanford's RegLab and Institute for Human-Centered AI. Read the scope carefully:
- It was the first preregistered empirical evaluation of AI-driven legal research tools, released in May 2024.
- It tested three systems: LexisNexis's Lexis+ AI, and Thomson Reuters's Westlaw AI-Assisted Research and Ask Practical Law AI. It compared them against GPT-4.
- The dataset was over 200 hand-built open-ended legal queries, scored by people.
- The paper's abstract states the tools "hallucinate between 17% and 33% of the time." Stanford's own summary of the work reports Lexis+ AI and Ask Practical Law AI producing incorrect information more than 17% of the time and Westlaw's AI-Assisted Research more than 34% of the time.
Two things follow. First, the study's finding was that vendor claims of eliminating hallucinations were overstated, while grounding did reduce errors substantially compared to a general chatbot. That conclusion still holds as a warning about marketing language. Second, and this is the part that gets dropped, the products tested no longer exist under those names or in those configurations. Westlaw AI-Assisted Research is a 2023 feature. Westlaw Advantage with Deep Research and Protege inside Lexis+ both shipped after the fieldwork. Quoting a 2024 number as a 2026 accuracy score for either platform is not evidence, it is a citation laundering exercise.
The honest position is that there is currently no independent, preregistered, public benchmark of the 2026 releases from either vendor. Until one exists, treat every accuracy claim, from vendors and from review sites alike, as unverified.
What does not require a benchmark is the professional obligation. ABA Formal Opinion 512 makes understanding a generative AI tool's tendency to produce inaccurate output part of the duty of competence, which means verification time belongs in your cost model regardless of which vendor you pick. We track what happens when it is skipped in our record of AI hallucination sanctions.
What does each cost in practice?
The pricing transparency gap between these two vendors is real and easy to verify.
Thomson Reuters publishes list rates for small firms. Its Westlaw Advantage plans and pricing page, checked on September 3, 2026, states that pricing starts at $256.75 per month for single-circuit coverage and $399.75 per month for all states and federal coverage, with 12% off a two-year subscription and 18% off a three-year subscription. The same page states that online pricing is available only to law firms with up to 10 attorneys, and that larger firms must contact sales. Those are content subscription rates. The AI bundles, Westlaw Advantage with CoCounsel Essentials, CoCounsel Essentials on its own, and CoCounsel Legal, are quoted rather than listed. We walk the arithmetic through a full first-year total in our CoCounsel pricing breakdown.
LexisNexis publishes nothing comparable. Its own FAQ says pricing varies by organization size, capabilities required and content scope, and directs buyers to contact sales for a custom quote. A free trial is offered. That is the entire published position, which is why our directory records Lexis+ AI and CoCounsel alike as pricing on request.
Practically, both vendors sell multi-year content subscriptions with the AI attached, so the number that matters is total contracted spend across the term, not a per-seat AI price. Any comparison that gives you a tidy monthly figure for "Lexis AI" is quoting someone's negotiated deal, not a rate card. Our legal AI pricing guide covers how the category prices generally.
Should AI make you switch platforms?
Usually not, and this is the part neither vendor will tell you.
Switching research platforms is not a software migration. It is a retraining exercise across every lawyer and paralegal who researches, a citation-format and workflow change, a knowledge-base reconnection, and the loss of whatever institutional muscle memory your team built in the incumbent system. Against that, the AI capability delta between the two current platforms is narrow, and it moves every quarter. A feature advantage you switch for in Q3 may be matched in Q1.
The cases where switching does pencil out are specific: your current contract is already ending and you are renegotiating anyway; your practice depends on content one publisher has and the other does not; or your firm has run a real side-by-side pilot on your own matters and found a consistent, repeatable quality gap. Vendor demos do not count as evidence, because demos are run on queries the vendor chose.
Scorecard: Westlaw Advantage vs Lexis+ with Protege
| Dimension | Westlaw Advantage (with CoCounsel) | Lexis+ with Protege |
|---|---|---|
| Grounding backbone | West editorial layer: headnotes, Key Number System, KeyCite | Shepard's treatment history, Practical Guidance, plus a Shepard's Verify pass over AI output |
| Drafting | Sold largely through CoCounsel Essentials and CoCounsel Legal rather than the research plan alone | Built into the platform, transactional and litigation drafting, DMS-connected |
| Bundling | AI is attached to content plans, CoCounsel rides on the Westlaw subscription | AI and content sold as one platform, no separately named AI SKU since the rename |
| Pricing transparency | List rates published for firms up to 10 attorneys, AI bundles quoted | No published rates at any firm size, quote only |
| Platform lock-in | High, West taxonomy and citation workflow become house standard | High, Shepard's workflow plus DMS and Vault integrations deepen over time |
| Independent accuracy evidence | None for the current release | None for the current release |
Read the last row as the honest summary. On evidence available to the public today, no one can tell you which of these two platforms is more accurate in 2026, and anyone who gives you a percentage is repeating a 2024 test of different software.
If you are choosing your first AI research tool rather than replacing an incumbent, the field is wider than these two. Our Harvey vs CoCounsel vs Legora comparison covers the platform-agnostic options, and our guide for small firms covers the tier below enterprise pricing.
This article is general information about legal software, not legal advice, and it does not create a lawyer-client relationship. Pricing and product names were verified on vendor sites on September 3, 2026, and both vendors revise both frequently.
Frequently asked questions
Which hallucinates less, Westlaw or Lexis AI?
Nobody can answer that for the products currently on sale. The only independent preregistered study, from Stanford's RegLab and HAI in May 2024, found hallucination rates between roughly 17% and 34% depending on the system, with Lexis+ AI and Ask Practical Law AI at the lower end and Westlaw AI-Assisted Research at the higher end. Those systems have since been replaced or rebuilt. Treat any 2026 percentage you see as unsourced unless it names a methodology, a query set, and a test date. Our framework for evaluating legal AI covers how to run your own test on your own matters, which is the only current way to get an answer you can rely on.
Is CoCounsel part of Westlaw now?
Effectively, yes. Thomson Reuters sells Westlaw Advantage with CoCounsel Essentials as a bundle, alongside CoCounsel Essentials on its own and CoCounsel Legal as the full stack. CoCounsel is not a standalone AI product you bolt onto a rival's research platform. It is grounded in Westlaw and Practical Law content, which is precisely why it is sold with them. CoCounsel came to Thomson Reuters through the 2023 Casetext acquisition, one of the deals in our legal AI acquisitions tracker.
What does Lexis+ AI cost?
There is no published price. LexisNexis states that pricing depends on organization size, the capabilities you need, and the scope of content access, and asks buyers to request a quote. A free trial is offered. Any specific per-seat figure circulating online is someone's reported deal terms rather than a rate card, so budget from your own quote and compare it against a total contracted spend rather than a monthly number.
Can small firms afford either?
Small firms can afford the research subscriptions, and Thomson Reuters is the only one of the two that lets you check that without a sales call, with published Westlaw Advantage rates for firms up to 10 attorneys. The AI layer on top is quoted separately in both cases, so the honest answer is that you cannot know your all-in cost without talking to a rep. Many small firms get better value from a purpose-built AI tool alongside a cheaper research subscription than from a premium bundle they use at 20% capacity. Run the numbers against actual research hours before signing a three-year term.
Browse every tool we track in the category on our legal research directory page, where listings are free, ownership changes are tracked, and featured placement is labeled and never affects the ranking.