Legal Research Software in Legal (2026)

What legal research software covers

Legal research software finds and checks the law: case law, statutes, regulations and the commentary around them, usually with a citation check that shows whether a case is still good law. The table mixes several kinds of product.

Research databases with AI added. Westlaw, Lexis+ with Protégé, Bloomberg Law and Vincent by Clio (formerly vLex Vincent AI) pair a database of primary and secondary law with an AI assistant. CoCounsel Legal is Thomson Reuters's assistant, which Thomson Reuters says reasons from an organization's own knowledge, Practical Law guidance and Westlaw authority.

General legal AI platforms. Harvey and Legora offer research next to drafting and review. Eudia's site describes research "from the law and your own documents" with sentence-level citations and lists a U.S. case-law source. Alexi generates research memos, and Paxton researches authority and drafts court-formatted documents.

Specialist and narrower-scope tools. Jus AI covers arbitration and international law. Midpage and Descrybe cover US case law, list monthly prices, and run inside assistants such as Claude and ChatGPT.

What the documented deployments report

Thomson Reuters published all four CoCounsel records on its own site.

Hawaii Disability Legal Services appears in a December 2025 blog post as a quote from attorney Diane Haar: "The time savings is massive, and time is money." She gives no figure and adds, "You still have to read the cases" (case study).

Valiant Law (October 2024) also gives no figure. Attorney Safa Riadh says he used the tool at trial and the judge agreed the issue was up to the judge's discretion. (case study).

Thomas & Thomas (December 2025) quotes attorney Nolan Thomas: the tools "allow us to manage 10 more clients a month, so the increased revenue more than justifies the expense." (case study).

Legal Aid Society of San Bernardino reports twice as many urgent hotline calls handled, per-call time cut from 30 minutes to 15, and 20% more cases onboarded. Its "up to 90% faster" concerns responses to opposing counsel, and the 50% saving is one attorney's remark about "a recent case" (case study).

Clayton Utz is a June 2024 trade-press report (IDM) on its adoption of Lexis+ AI. The 4.5 hours cut to 30 minutes belongs to a lawyer at Holding Redlich, another firm, on one task. (case study).

Irwin Mitchell is a LexisNexis announcement (May 24, 2024), reposted by a trade site, that the firm has selected Lexis+ AI, written in the future tense. Its figures come from a November 2023 survey of early US users, 74% of whom estimated they would save 7 hours a week on research (case study).

Cuatrecasas is Harvey's customer page. It says more than 80% of the firm's lawyers use Harvey daily, "generating over 3,300 prompts" with no period given, and describes a research tool built on more than 3,000 of the firm's own templates, briefs and memos. It gives no time or accuracy figure; a partner says it is "not about saving me time, but rather improving the outcome of my work" (case study).

White & Case is a January 2026 news report (IT Brief Asia) of an announced Legora rollout to all lawyers in 43 offices. It reports plans and Legora's company figures, not results (case study).

Eudia's own page on Cargill reports about 50% less "contracting research time" from querying Cargill's own contract data. It gives no baseline, and the work is search over the company's contracts, not case-law research (case study).

Jus Mundi has four records but is listed without a rank because its vendor profile on this site is not published. All four are articles on Daily Jus, Jus Mundi's own publication, each with a different measure. Dunning Vallejo & MacDonald says Jus AI saves "50 to 70% of the time" it would usually spend, "in many cases—or most cases, I'd say." For one 48-hour deadline it describes "a couple hours tweaking the prompt" and then "two hours and 15 minutes or less", against an estimated 20 to 30 hours (case study). Duan & Duan reports arbitrator research falling from "a week or two" to "a day or two" (case study), Templars two to three days to 15 to 30 minutes (case study), and Wordstone one to two weeks to three minutes (case study).

Two records name no vendor. The Decoder (July 21, 2026) reports a randomized field experiment, by researchers at ETH Zurich, Imperial College London and the New Economic School, of 1,559 judges in 118 Pakistani courts using a GPT-4 assistant. Judges who also had six 90-minute lectures used it about four times as much as judges given access and a general seminar; The Decoder writes that "AI access alone did little" (case study). A GitHub page in Vectara's repository summarizes Mata v. Avianca (2023): a lawyer used ChatGPT for research, none of the six cases cited in the brief existed, and ChatGPT told him they were real when he asked (case study). One incident shows what an unchecked chatbot can do, not how often.

What this evidence cannot tell you

Vendors wrote most of it. As of November 2026, at least 21 of the 30 records were written or published by the vendor: 16 are on the vendor's own website, 4 are Jus Mundi's Daily Jus articles, and 1 is the LexisNexis announcement reposted by a trade site. Seven records name no vendor, and 22 carry no source publication date. The ranking counts published records, not quality or market share.

It cannot say whether Westlaw or LexisNexis is better. No record is about Westlaw alone. Three of the four CoCounsel records describe Westlaw's AI-Assisted Research, and the fourth describes CoCounsel working with Westlaw and Practical Law content. LexisNexis has two records. None compares research products on the same questions.

A 2024 Stanford study did. Stanford RegLab and HAI researchers tested Lexis+ AI, Ask Practical Law AI and Westlaw AI-Assisted Research on benchmark queries. The tools reduced errors compared with general-purpose models like GPT-4 but still hallucinated: the first two produced incorrect information more than 17% of the time, Westlaw AI-Assisted Research more than 34%. An earlier Stanford study found general-purpose chatbots hallucinated 58% to 82% of the time on legal queries. These are 2024 results on one benchmark.

Figures have different scopes. They cover one task, one firm's estimate, or other users' survey answers. None of the vendor-written sources above reports how often a research answer was wrong.

What the listed vendors publish about price

As of November 6, 2026, five of the 13 vendors in the table show a dollar price in their rows. The units differ, so the figures do not compare directly.

Westlaw Advantage starts at $256.75 per month for single-circuit coverage and $399.75 for all states and federal, with 12% off a two-year term and 18% off three years (pricing page). Online pricing is limited to new-customer firms of up to 10 attorneys; checkout asks for the number of attorneys, and the page does not say if the price is per attorney. Thomson Reuters's research page shows two starting prices for Westlaw Advantage, $256.75 and $288.60 per month, and $551.85 per month for CoCounsel Legal, with no buyer segment or per-attorney unit (pricing page).

Paxton's individual plan lists $499 per user per month or $2,999 per user per year after a 7-day free trial (pricing page). Midpage lists $100 per seat per month billed monthly, $25 per month for US law school students and faculty with proof, and two free weeks (pricing page). Descrybe lists $50 per month for its platform and $25 for Legal Engine alone, which it says is flat with no per-search charges, and says subscriptions to the assistants it connects to are bought separately (pricing page).

The other eight (LexisNexis, Bloomberg Law, Harvey, Legora, Alexi, Vincent by Clio, Jus Mundi and Eudia) show no figure.

Questions to ask in a demo

• Run a question from our practice area and jurisdiction. Which sources did the answer draw on?
• What does the quoted time saving measure, and against what baseline?
• How often are answers wrong or citations unsupported on a test set we choose? What does the tool show when it finds nothing?
• Which citator checks the authorities the AI cites?
• What is the price unit (user, seat, firm, coverage tier), and what does the quote become at renewal? Thomson Reuters's online prices are for new customers only.
• What training do you provide? In one trial of Pakistani judges, access alone produced much less use than access plus training.
• Which reference customers are independent of you, and who wrote their published results?

How to use the ranking

Start with the legal research use case and the records nearest your practice. Check who wrote each cited page and what its figure measures, then put these questions to every vendor. For document review across a case file, see e-discovery software; for contracts, see contract lifecycle management. Other categories are under software.