Will AI replace lawyers? Federal job data still projects growth, and independent tests show legal AI errs too often to work unsupervised. Here is the evidence.
Will AI replace lawyers? Not on the evidence available today. The federal government still projects more lawyer jobs over the next decade, and the best independent tests of legal AI show error rates that make unsupervised use risky. What AI is changing is the work inside a lawyer’s day, and that shift is already showing up in how firms bill and staff matters.
This piece looks at three kinds of evidence: labor projections, independent accuracy testing, and what the professional rules say about who is responsible for the work.
What the Labor Data Says
The US Bureau of Labor Statistics Occupational Outlook Handbook counts about 863,700 lawyer jobs in 2025 and projects employment to grow 5% from 2025 to 2035, faster than the 3% average for all occupations. It expects about 28,700 openings a year, most of them to replace lawyers who retire or change careers.
The BLS addresses automation directly. It says some routine legal work may be automated, but that this efficiency “is not expected to reduce overall demand for lawyers,” and that automation is expected to give lawyers more time for client meetings and strategy. That is a projection, not a guarantee, and it was last updated on August 27, 2026, before the newest agentic products had much time in the market. Still, it is the most authoritative public forecast available, and it points toward change in tasks rather than fewer lawyers.
What Independent Tests Show
The strongest public evidence on accuracy comes from a preregistered study by researchers at Stanford, published in 2025 in the Journal of Empirical Legal Studies. It ran more than 200 legal queries through AI research tools from LexisNexis and Thomson Reuters and found that each hallucinated between 17% and 33% of the time. The best performer answered 65% of queries accurately. The tools did better than a general-purpose chatbot, but none came close to error-free.
Contract work tells a similar story. On Ivo’s ContractBench, which scores whether AI models make the same calls experienced contracts lawyers would, no model cleared 60%, as we covered in our report on the benchmark. Courts are also seeing the cost of skipped checks: our piece on two federal judges putting a price on unchecked AI citations shows how fabricated citations turn into sanctions.
Both tests used products that have since been updated, so the numbers describe a moment, not a ceiling. They do show why vendors sell these tools as assistants that need a lawyer’s review.
Who Answers for the Work, and How It Is Shifting
The professional rules place responsibility on the lawyer, whatever tool produced the draft. ABA Model Rule 1.1, Comment 8, says lawyers should keep abreast of “the benefits and risks associated with relevant technology.” ABA Formal Opinion 512, explained in our guide, applies the existing duties of competence, confidentiality, supervision and reasonable fees to generative AI. Our roundup of state bar AI opinions shows bars taking the same line.
As long as a licensed person answers for the work, AI changes who does the first draft, not who signs it. That is a description of the rules, not legal advice for any particular practice.
The pressure is on tasks and pricing. Data from 8am, covered in our piece on billable hour trends, shows the hourly model still dominates at small and midsize firms, even as tools make some tasks much faster. Vendors are now building products that log the time lawyers spend working in AI, as we reported on Legora and Laurel. If first drafts take minutes, the open question is how firms price the review, judgment and accountability that remain.
The roles most exposed are those built on volume: first-pass document review, routine drafting and research summaries. The least exposed involve judgment, negotiation, advocacy and client trust. That split is a reading of where the tools work today, not a forecast of any job.
Common Questions
Will AI replace lawyers in 10 years?
The BLS projects lawyer employment to grow 5% from 2025 to 2035. No authoritative forecast we found projects a decline.
Can AI replace lawyers for simple legal tasks?
AI tools can draft and summarize, but independent tests show meaningful error rates, and the professional rules keep a licensed lawyer responsible for the result.
Which legal jobs are most affected by AI?
Work built on volume, such as first-pass review and routine drafting, is where tools are most capable today.
Is legal AI accurate enough to use without review?
The Stanford study found hallucination rates of 17% to 33% in leading research tools. That supports review, not unsupervised use.
The Verdict
Fluff. The claim that AI will replace lawyers is not supported by federal labor projections or by independent accuracy testing. The real change is narrower and still significant: fewer hours on first drafts, more on checking and judgment, and growing pressure on how that work is billed.
In this story
See every round in the Funding Tracker