Guides
The State of AI in Legal Practice in 2026: Every Number That Matters
Legal AI statistics for 2026, dated and sourced: adoption, client pressure, funding, prices, juniors, burnout, hallucination cases and regulation, with caveats.
On this page
- How to read this page
- 1. Adoption: who uses what
- 2. Value and client pressure
- 3. Money: vendors and funding
- 4. Prices, as of September 2026
- 5. People: hiring, juniors and burnout
- 6. Risk: hallucinations and sanctions
- 7. Regulation in numbers
- What the numbers do not tell you
- How to use these numbers in your firm
As of September 2026, 94% of law firms use or are exploring generative AI, 79% of legal professionals use it personally, 74% of professionals use it several times a week, and courts have found hallucinated material in 2,041 decisions. Those four figures come from four surveys with four different denominators. This page collects the numbers that matter, with date, source and what each can support.
How to read this page
Most legal AI statistics come from parties with a product to sell: Clio, Thomson Reuters and LexisNexis survey their own audiences; Harvey and CMS publish their own pilot results. "Use AI" spans everything from one ChatGPT query to daily agent runs. Firm-level and individual-level figures diverge sharply. And vendor metrics change definition: Harvey's "3,000+ customers" (September 2026) follows "1,300+ organisations" (March 2026), which is a new way of counting, not a doubling. Where two credible sources disagree, both appear below. Widely repeated numbers that trace to a panel anecdote rather than a study, such as the Legalweek 2026 claim that diligence fell from 15–20 associate hours to two, are left out.
1. Adoption: who uses what
| Statistic | Figure | Source and date |
|---|---|---|
| Firms using or exploring generative AI | 94% (80% in 2025) | ILTA 2026 Technology Survey, 500+ firms, 140,000 lawyers, 14 Sep 2026 |
| Products in use at firms | Microsoft 365 Copilot 76%; CoCounsel 44%; Claude 44%; Harvey 43%; Lexis+ AI 26%; ChatGPT Enterprise 22%; Legora 16% | ILTA 2026 |
| Harvey among firms with 700+ lawyers | 84% | ILTA 2026 |
| Legal professionals using AI | 79% (19% in 2023); large firms 87%; solos 71% | Clio Legal Trends Report 2025, autumn 2025 |
| Use of legal-specific AI tools | 40%, down from 58% | Clio 2025 |
| Professionals using AI several times a week / multiple times a day | 74% / 44% | Thomson Reuters Future of Professionals 2026, 1,816 respondents, 62 countries, Jun 2026 |
| Using unsanctioned "shadow AI" | 34% | Thomson Reuters 2026 |
| Enterprise-wide generative AI in firms and departments | 43% (14% in 2024) | Thomson Reuters Institute, 9 Apr 2026 |
| US firm lawyers using AI weekly / three or more times a week | 70% / 47% (27% in 2025) | Law360 Pulse, 506 lawyers, Nov 2025–Jan 2026 |
| Federal judges using at least one AI tool | 61.6% of 112; Westlaw AI 38.4%, ChatGPT 28.6% | Northwestern/NYC Bar survey, 27 Mar 2026 |
| Firm has no AI policy, or lawyer does not know | 53% | Clio 2025 |
Sources: LawSites on ILTA 2026, Clio 2025 summary, Thomson Reuters Future of Professionals 2026, Texas Bar 2026 AI survey.
What it means: adoption is individual and informal ahead of institutional and governed. The ILTA ranking shows what firms have licensed; the Clio numbers show what lawyers actually open, and it is increasingly a general chatbot. The gap between 94% of firms "exploring" and 43% with enterprise-wide deployment is where most of the risk in section 6 lives. A fuller answer to the headline question is on what percentage of lawyers use AI.
2. Value and client pressure
| Statistic | Figure | Source and date |
|---|---|---|
| Professionals who say their organisation falls short of AI's potential | 91% | Thomson Reuters 2026 |
| AI meets expectations, with vs without a named strategy | 66% vs 22% | Thomson Reuters 2026 |
| Estimated time saved per professional per year | 240 hours (~$19,000 of value) | Thomson Reuters 2026, self-reported |
| Corporate clients calling AI-enabled improvements very important / saying providers deliver / reconsidering relationships | 78% / 6% / 32% | Thomson Reuters 2026 |
| US in-house counsel who started using generative AI in 2025 / unaware whether outside firms use it / see no savings | 52% / 59% / 59% | ACC and Everlaw, 2025 |
| Legal departments piloting agentic AI / fully embedded / no adoption | 61% / 10% / 2% (76% in 2024) | Deloitte, Apr–May 2026 |
| In-house teams that use, optimise and measure AI at scale / cannot show spend paid off / expect or negotiate AI rate cuts | 7% / 83% / 92% | Axiom, 528 leaders, 9 Jul 2026 |
| Firms saying AI improved profitability per matter / partners uncomfortable discussing AI with clients | 31% / 35% | BigHand, 16 Sep 2026 |
| Clients wanting disclosure / lawyers who rarely disclose / clients less likely to trust AI-using lawyers | 78% / 35% / 36% | Clio 2025 |
| Growth in demand for Big Law hours, H1 2026 | +4.2% (1.5% norm) | Citi, Sep 2026 |
| Measurable benefit within three months / hours saved by power users | 68% / ~37 hours a month | RSGI independent study of Harvey, 40 organisations, 19 Nov 2025 |
| CMS Harvey pilot | 93% report productivity gains; up to 117.9 hours per lawyer per year | CMS, 10 Dec 2025 |
Sources: ACC/Everlaw, Axiom, Artificial Lawyer on BigHand, Bloomberg Law on the Citi data.
What it means: the value is real where it is measured (RSGI, CMS) and invisible where it is not (Axiom's 83%). Clients have noticed. Citi now asks firms bidding for its work to document AI savings, Morgan Stanley is expanding competitive bidding, and Meta's outside-counsel guidelines refuse payment for tasks AI could have done (March 2026). Yet demand for hours rose. How firms reconcile those facts is the subject of AI and the billable hour and, from the buyer's side, AI for in-house counsel.
3. Money: vendors and funding
| Statistic | Figure | Source and date |
|---|---|---|
| Harvey | $550m raised at $15.5bn (9 Sep 2026); ARR above $400m; 3,000+ customers; 80% of the top 100 law firms (vendor-reported); over $1.5bn raised since 2022 | LawSites, 9 Sep 2026 |
| Harvey valuation path | $3bn (Feb 2025) → $5bn (Jun 2025) → $8bn (late 2025) → $11bn (25 Mar 2026) → $15.5bn | Harvey, CNBC, LawSites |
| Legora | $550m Series D at $5.55bn (10 Mar 2026); $50m extension at $5.6bn (Apr 2026); ARR $100m+, reportedly ~$150m; 1,480+ customers; $800m raised in total; three acquisitions in 2026 | Legora, TechCrunch, 30 Apr 2026 |
| Clio | $1bn acquisition of vLex closed; $500m Series G at $5bn; ~$400m ARR | ABA Journal, 10 Nov 2025 |
| Legal-tech funding 2025 | $5.99bn (+22%) across 292 companies (−27%); exits $2.29bn | Artificial Lawyer, 6 Jan 2026 |
| Market reaction to Anthropic's legal plugin, 3 Feb 2026 | Thomson Reuters −18% intraday; RELX −14% (steepest since 1988); Wolters Kluwer −13% | Legal.io, Morningstar |
| Big Law building its own | Kirkland & Ellis $500m platform (May–Jun 2026); Latham & Watkins own Nvidia H200 servers (Sep 2026) | Global Legal Post, Bloomberg Law |
| Cautionary tale | Robin AI failed to close a $50m round and sought a buyer (Oct 2025); team acqui-hired by Microsoft (Jan 2026) | Artificial Lawyer |
Sources: LawSites on Harvey's round, TechCrunch on Legora, Artificial Lawyer on 2025 funding, Legal.io on the February sell-off.
What it means: capital is concentrating in two platforms while the number of funded companies falls, and the frontier labs now sell to firms directly (Claude for Legal, 12 May 2026; Astra for Law, 17 September 2026). Harvey at roughly 39 times ARR is priced to remain the interface; the February sell-off priced the possibility that it does not. Reviews of the two leaders are at Harvey and Legora.
4. Prices, as of September 2026
Published or reliably reported price points, for orientation rather than procurement: Harvey about $1,000–1,200 per seat per month with a 20-seat minimum (secondary sources; no published price); Legora from about $3,000 per user per year with a 10-seat minimum (secondary); CoCounsel Legal $104–$639 per user per month (published); Clio Work $199; Spellbook roughly $99–$350; GC AI $500; Microsoft 365 Copilot $30; Claude Team $25 or $125 per seat; ChatGPT Business $25 monthly or $20 annual; AI:ssociate (Austria) from €39. The full table with dates and caveats is the legal AI pricing guide.
5. People: hiring, juniors and burnout
| Statistic | Figure | Source and date |
|---|---|---|
| Entry-level hires at firms with 500+ lawyers, Class of 2025 | 6,588, down 538 (−7.5%); first fall since 2011 | NALP, Aug 2026 |
| Large firms predicting fewer first-year associates by 2028 | 67% | Citi survey of 57 firms, Sep 2026 |
| UK training-contract places for 2027 | −1.7% (2,593 to 2,549); Linklaters 100 to 60; Freshfields 85 to 70 | Legal Cheek, 10 Sep 2026 |
| Lawyers screening positive for high burnout / severe / considered leaving | 47.4% / 15.6% / 27% | ABA and Krill, 37,000 lawyers, 10 Sep 2026 |
| Professionals worried about erosion of independent judgment / would refuse a job without professional-grade AI | 48% / 24% | Thomson Reuters 2026 |
| UK junior lawyers: faster with AI / say deep reasoning is the biggest gap / think AI strengthens learning | 58% / 72% / 2% | LexisNexis, ~900 lawyers, Feb 2026 |
| Heavy users who feel positive about AI | 51%, down from 73% | Law360, 31 Mar 2026 |
Sources: ABA burnout study, Legal Cheek on trainee numbers, Bloomberg Law on first-year hiring.
What it means: the hiring numbers are small so far and confounded by the economy, but the direction is consistent. Firms are answering with structure rather than headcount: Ropes & Gray gives juniors 20% of billable hours for AI work; Sidley added a compulsory AI seat for London trainees (September 2026). Wolf Theiss's Chief Digital & Technology Officer Leonhard Wassiq has argued that firms hiring fewer people because of AI "will be hit hard". The debate is set out in junior lawyers and AI.
6. Risk: hallucinations and sanctions
| Statistic | Figure | Source and date |
|---|---|---|
| Court decisions worldwide involving hallucinated AI material | 2,041 (USA 1,395; Canada 217; Australia 111; UK 69; Germany 11) | Charlotin database, 14 Sep 2026 |
| By actor | Pro se litigants 1,174; lawyers 811; judges 32; experts 15 | Charlotin |
| Growth | ~200 (mid-2025) → 712 (29 Dec 2025) → 1,227 (Apr 2026) → 1,598 (9 Jun 2026) → 2,041; roughly eight a day | Charlotin, Bloomberg Law |
| Largest US sanction | $110,204.38, Couvrette v. Wisnovsky (D. Or., 12 Dec 2025) | ABA Journal |
| US penalties, Q1 2026 | Over $145,000 | HAQQ / ComplexDiscovery |
| First appellate fines | $15,000 per attorney plus fees and double costs, Whiting v. City of Athens (6th Cir., 13 Mar 2026) | Norton Rose Fulbright |
| Bar consequences | Nebraska interim suspension, 57 of 63 citations defective (Apr 2026); SRA v Kumar, first striking-off for AI hallucinations (Aug 2026) | Vaquill, Legal Futures |
| Paid legal tools in older independent testing | Lexis+ AI 17%, Westlaw AI-Assisted Research 33% hallucination | Stanford RegLab, 2024 products |
| Answers citing or misapplying law that did not support the claim | 24% of 3,000 | HAQQ benchmark, 5 Jun 2026 |
Sources: Charlotin database, ABA Journal on Couvrette, Norton Rose Fulbright sanctions update.
What it means: lawyers are 40% of the database but nearly all of the fines, and the firms involved include Morgan & Morgan, Butler Snow, Sullivan & Cromwell and Pinsent Masons, most of which had AI policies. Candour mitigates; denial aggravates. The mechanism is explained in AI hallucinations in legal practice; the case list is at AI hallucination sanctions.
7. Regulation in numbers
Seventeen US jurisdictions plus the ABA have formal AI ethics opinions as of June 2026; only Alabama added one in 2026 (Legal AI Compliance tracker). Over 30 of the 94 federal district courts have AI disclosure or certification requirements, and separately more than 300 individual federal judges have standing orders, a different unit that is often conflated (AI court disclosure map, Apr 2026). Forty states have adopted the Comment 8 duty of technology competence. In the EU, the Article 4 AI-literacy duty has applied since 2 February 2025 and was reworded but not deferred by the Digital Omnibus; Article 50 transparency duties apply from 2 August 2026; Annex III high-risk obligations moved to 2 December 2027. The SRA logged 42 AI-misuse reports in a year. No major US malpractice carrier had filed an AI exclusion as of 5 May 2026. The maps are at US state bar opinions, court standing orders and the EU AI Act for lawyers.
What the numbers do not tell you
First, nobody has measured output quality at scale outside vendor benchmarks, which disagree by a factor of 30 depending on scoring rules; see legal AI benchmarks explained. Second, "time saved" is self-reported and rarely reconciled with hours billed, which is why 240 hours saved per professional coexists with a 4.2% rise in hours demanded. Third, the surveys under-sample solos and firms of two to 25 lawyers, who make up the majority of sanctioned lawyers and are the least likely to have a policy.
How to use these numbers in your firm
Take three into the next partners' meeting. The policy gap (53% report no policy or do not know of one) justifies writing one this quarter; the AI policy page has the structure. The strategy gap (66% versus 22% satisfaction) justifies naming an owner. The training gap (54% of firms with none, against an EU AI-literacy duty that already applies) justifies a programme, whether internal or external; the AI literacy page sets out what Article 4 expects. Closing the distance between the adoption numbers in section 1 and the value numbers in section 2 is what the AI Lab for Lawyers course is built for, with its final module covering the risk and regulation figures above. Every figure here carries a date because most will have moved by the end of the year.
Frequently asked questions
What percentage of lawyers use AI in 2026?
It depends on the survey and the definition. Clio's 2025 Legal Trends Report found 79% of legal professionals use AI in some form, 87% in large firms and 71% of solos. ILTA's September 2026 survey found 94% of firms use or are exploring generative AI. Thomson Reuters found 74% of professionals use it several times a week. Read each with its denominator: firms, individuals or frequent users.
Which AI tools do law firms use most in 2026?
ILTA's 2026 Technology Survey of 500-plus firms puts Microsoft 365 Copilot first at 76%, then CoCounsel and Claude at 44% each, Harvey at 43%, Lexis+ AI at 26%, ChatGPT Enterprise at 22% and Legora at 16%. Among firms with more than 700 lawyers Harvey reaches 84%. Clio separately found that use of legal-specific tools fell to 40% as lawyers defaulted to general chatbots.
How many lawyers have been sanctioned for AI hallucinations?
Damien Charlotin's database listed 2,041 court decisions worldwide involving hallucinated AI material as of 14 September 2026, of which 811 involved lawyers, 1,174 self-represented litigants and 32 judges. Not every decision imposed a sanction. The largest US penalty to date is $110,204 in Couvrette v. Wisnovsky; US courts imposed more than $145,000 in the first quarter of 2026 alone.
Is legal AI actually saving law firms money?
The evidence is mixed. BigHand found 31% of firms say AI improved profitability per matter; Clio found 69% of wide adopters report revenue gains; CMS reported up to 117.9 hours saved per lawyer per year in its Harvey pilot. Yet Citi found demand for Big Law hours rose 4.2% in the first half of 2026, 59% of in-house counsel see no savings from outside firms, and 83% of legal departments cannot show whether their own AI spend paid off.
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
Questions
What Percentage of Lawyers Use AI in 2026? The Survey Numbers
The percentage of lawyers using AI in 2026 runs from 74% to 94% depending on the survey; daily use is 21–44% and only about 40% use legal-specific tools. Every number, dated.
6 min read
Guides
What Legal AI Really Costs in 2026: Harvey, Legora, CoCounsel, Claude and ChatGPT
Legal AI cost as of September 2026, from €39 to $1,200 per seat per month, with Harvey, Legora, CoCounsel, Claude and ChatGPT prices and worked budgets for 5 and 50 lawyers.
10 min read
Guides
AI and the Billable Hour: Who Keeps the AI Dividend?
AI and the billable hour in 2026: associate rates at $798 and hours up 4.2% while Citi demands AI savings. What the ethics rules allow, three fee scenarios and a clear view.
9 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
Guides
Junior Lawyers and AI: Training, Hiring and the Missing Rung
Junior lawyers and AI in 2026: Big Law hired 538 fewer graduates, 2% of UK juniors say AI helps them learn, and firms are rebuilding training with AI seats.
11 min read
Tools
Harvey AI in 2026: What It Does, What It Costs and Who Uses It
Harvey AI reviewed as of September 2026. The $15.5bn valuation, Harvey II, Tenet, 500-plus agents, who uses it, estimated seat prices and its own benchmark.
9 min read