MMOOZZGG.com

AI for lawyers

AI in law statistics: what the main studies show

6 min read6 sources

AI in law statistics for 2023–2026 agree on two points: lawyers already work with AI in large numbers, and errors of legal AI tools remain visible. According to the Clio Legal Trends 2025 report, 79% of lawyers use AI. Below are the main studies with their numbers, methods and limits.

On this page · 5 sections

How many lawyers use AI

Clio, Legal Trends 2025 report: 79% of lawyers use AI, the same as in 2024. In large firms the share is higher, 87%, and among solo practitioners it is 71%.

The gap between generations is clear. 28% of millennial lawyers use AI widely in their work, and only 5% of baby boomers, so lawyers with very different practices work side by side in one firm.

The share of general tools is growing. 40% of lawyers use specialized legal AI tools, against 58% in 2024; the report’s authors link this to the free plans of general chatbots, which are easier to start using.

Rules lag behind practice. 53% of lawyers say their firm has no AI policy or they do not know about it. Among the policies that exist, 30% allow and encourage AI, 12% allow it without encouragement, and 5% forbid it.

Expectations run ahead of adoption. In a 2025 Thomson Reuters survey, 80% of professionals believe that AI will change their profession greatly or fundamentally within 5 years, but only 38% expect significant change in their own organization this year. 46% note a skills gap in their team, mostly in technology and data.

This gives a firm its first practical step: an AI use policy that names the allowed tools, tasks and the procedure for checking results.

Data limit: this is a survey outside Russia, so its numbers apply to the Russian market as a benchmark, not as a measurement.

Time savings

Thomson Reuters, Future of Professionals 2025: a survey of 2,275 professionals in law, tax, trade, accounting, risk and compliance ran in February and March 2025. Respondents expect AI to save a professional 5 hours a week in the coming year; the 2024 forecast was 4 hours.

In money this is on average $19,000 a year per professional, and $32 billion a year for US lawyers and accountants together. This is the respondents’ forecast, not a measurement of actual savings.

The return depends more on strategy than on the tool. Only 22% of organizations have a visible AI strategy, and they are twice as likely to see revenue growth from AI. 53% of professionals say their organization already gets a return from AI, and 88% want an AI assistant for their profession.

Clio gives a similar picture from the users’ side: 65% of lawyers who work with AI note better quality of work, and 63% faster answers to clients. 62% say AI cut routine work, and 58% that it improved accuracy.

There is a money effect too. 36% of lawyers note a positive effect of AI on revenue, and among those who adopted AI widely, 69%. The difference does not prove the cause, but it matches the Thomson Reuters finding on strategy: systematic adoption gives more than one-off requests to a chatbot. Almost half of lawyers, 47%, say AI reduced stress.

How many hours this gives your firm can be estimated in the time savings calculator: it calculates from the number of lawyers, the hourly rate and the share of routine work.

Errors of legal AI

Stanford RegLab and HAI, May 2024: the researchers prepared and pre-registered a set of more than 200 open legal questions in advance and tested legal AI research tools on it.

The result: Lexis+ AI and Ask Practical Law AI made errors in more than 17% of answers, and Westlaw AI-Assisted Research in more than 34%. In an earlier study by the same authors, general chatbots erred on 58–82% of legal questions.

The authors separate two types of error. An answer can describe the law wrongly, or it can describe it correctly but cite a source that does not support what it says. The second type is more dangerous: the citation looks real, and a lawyer trusts it without checking.

Errors also go beyond the labs. The database of AI hallucination cases in courts kept by Damien Charlotin counted 2,190 cases as of 15 October 2026: the US 1,492, Canada 232, Australia 112, the UK 76, Israel 58, and Russia 1.

Searching a source database before answering lowers the number of errors compared with GPT-4, but does not remove them. So every citation is checked against the original source; how it ends in courts is covered on the page about AI hallucinations in courts.

Limit of the study: it tested the versions of the tools at the time of the study and questions on US law. The figures show the scale of the risk, not the error rate of a specific tool today.

The bar exam: GPT-4

OpenAI stated that GPT-4 passed the US bar exam at about the 90th percentile. Eric Martinez recalculated the result in the journal Artificial Intelligence and Law (2024): below the 69th percentile, and about the 48th on essays.

The recalculation shows how much the figure depends on the method. The same result looks like the 90th percentile or below the 69th, depending on which group of test takers you compare it with.

For practice, something else matters more. The model was weakest on essays, that is, on written argument, which is closest to a lawyer’s real work. Passing a test and running a case are not the same: an AI draft is useful, but the lawyer builds and signs the position in the case.

The practical conclusion for a firm: judge AI not by exams and advertising figures but on your own tasks. A pilot on one process, with a lawyer checking every answer, gives the firm its own statistics on errors and saved hours.

Labor market forecasts

Goldman Sachs, March 2023: generative AI could expose the equivalent of 300 million full-time jobs to automation. In the US, about a quarter of work tasks could be automated.

For legal occupations the estimate is above average: 44% of tasks, and 46% for administrative ones. This is a share of tasks, not of jobs: it is about what part of the work can be handed to AI, not about how many lawyers will no longer be needed.

Read the estimate with its date. Goldman Sachs made it in March 2023, before the Clio and Thomson Reuters surveys in the sections above, so it describes the technical potential for automation, not the actual replacement of tasks in firms.

The figures describe potential, not a plan. The firm decides which tasks to hand to AI, while decisions, signatures and negotiations stay with lawyers.

The question of the profession’s future is covered separately on the page “Will AI replace lawyers”: what the studies say and what is already changing in a firm’s work.

Sources and fact-checking

Checked by the editors on 16 October 2026.

  1. 2Civility (review of the Clio report), 2025 (2civility.org)
  2. Thomson Reuters, 26.06.2025 (thomsonreuters.com)
  3. Stanford HAI, 23.05.2024 (hai.stanford.edu)
  4. MIT DSpace, 30.03.2024 (dspace.mit.edu)
  5. Goldman Sachs Global Investment Research, 27.03.2023 (gspublishing.com)
  6. AI Hallucination Cases Database, 15.10.2026 (damiencharlotin.com)

Found a mistake? Tell us

Publications

Contact us