Will AI replace lawyers? What the data says in 2026

No, AI is not going to replace lawyers, but it is already doing a growing share of legal work, especially research and first drafts. The U.S. Bureau of Labor Statistics (BLS) rates lawyers among the occupations most exposed to AI, yet it still projects lawyer jobs to grow 5% by 2035 while legal secretary jobs shrink and paralegal jobs stall.

The reason is simple. AI is good at the parts of law that look like reading and writing: finding cases, summarizing documents, drafting a contract. It cannot hold a law license, stand up in court or take responsibility for advice, and it still invents cases with total confidence. Below are the official numbers, what AI can and can’t do in legal work, which legal jobs are most at risk, and how to stay on the right side of the change.

The numbers: very high exposure, and still growing

In August 2026 the BLS published official AI exposure ratings for 831 occupations, alongside its job projections for 2025 to 2035. The ratings combine five research measures, including real-world AI usage data from Anthropic and Microsoft, and place each occupation in one of four groups from low to very high exposure.

Lawyers are in the very high group, and every legal occupation in the data is rated high or very high. But look at where the same agency expects the jobs to go over the next ten years:

Similar exposure, different outcomes. Lawyer jobs grow faster than the 3% average for all occupations, from 863,700 in 2025 to about 904,300 in 2035, with roughly 28,700 openings a year. The jobs built around typing, filing and preparing documents are the ones that stall or shrink. Legal secretaries and administrative assistants are projected to lose about 8,300 jobs.

The BLS spells out why in its outlook for lawyers: some routine legal work may be automated, but this is “not expected to reduce overall demand for lawyers.” Instead, it expects automation to give lawyers more time for meeting clients and planning strategy. Pay reflects that position. The median lawyer earned $159,670 in May 2025, more than three times the median for all workers.

Can AI replace lawyers? What it can and can’t do

The best evidence comes from studies of how people actually use AI, instead of guesses about what it might do.

Microsoft analyzed 200,000 anonymized conversations with its Copilot assistant and scored U.S. occupations on how much of their work AI was used for, and how well. Lawyers scored 0.18, higher than almost two in three of the 785 occupations in its published data. The legal activity people brought to AI most often was researching laws and precedents, and AI completed that task in about 92% of those conversations. It did just as well when people asked for advice on legal or regulatory matters, completing about 95% of them.

That matches what lawyers report: AI is genuinely useful for the first pass of research and writing. The problem is the second pass. AI tools still produce answers that look right and aren’t.

• Even legal AI tools make things up. Stanford and Yale researchers tested the AI research tools sold by LexisNexis and Thomson Reuters, the companies behind Lexis and Westlaw. Each tool hallucinated, giving answers that were false or cited sources that didn’t support them, between 17% and 33% of the time. A general chatbot did worse.
• Courts keep catching it. In 2023 a federal judge in New York fined two lawyers and their firm $5,000 for filing a brief citing cases that ChatGPT had invented. That was the first famous case, not the last. A public database maintained by researcher Damien Charlotin now lists more than 2,200 court decisions worldwide dealing with AI-hallucinated material in filings.

So AI can draft the memo, but a lawyer has to check every case it cites, and is personally on the hook if they don’t. Here is how legal work splits:

• Most exposed: research, document review and first drafts. Finding relevant case law, summarizing depositions, reviewing contracts in due diligence, drafting routine letters and agreements. AI does a fast first version; a person must verify it.
• Partly exposed: briefs, pleadings and complex contracts. AI speeds up the drafting, but the argument, the strategy and the risk judgment are still the lawyer’s.
• Least exposed: advice, negotiation and advocacy. Telling a client what to do, negotiating a deal, cross-examining a witness, persuading a judge or jury. These rest on judgment, trust and a license.

Why AI can’t legally replace a lawyer

There is a hard limit that most “AI lawyer” headlines skip: practicing law requires a license. In the U.S., lawyers need a law degree and a state license, usually earned by passing a bar exam. With narrow exceptions, only a licensed lawyer can give legal advice to someone else or represent them in court, and software can’t hold that license.

Regulators enforce the line. In February 2025 the Federal Trade Commission finalized an order against DoNotPay, which had marketed itself as “the world’s first robot lawyer.” The FTC said the service didn’t live up to its claims of replacing a human lawyer’s expertise, ordered it to pay $193,000, and barred it from claiming it performs like a real lawyer without evidence.

The profession’s own rules point the same way. In July 2024 the American Bar Association issued Formal Opinion 512, its first ethics guidance on generative AI. Lawyers who use AI still owe clients competence, confidentiality and honest communication. They can bill for the time spent using an AI tool and reviewing its output, but not for learning the tool. In other words, AI changes how lawyers work, but the responsibility stays with the lawyer.

Will AI replace paralegals?

Paralegals face more pressure than lawyers, because more of their week is research, document preparation and organizing files. In its outlook, the BLS says AI is expected to make paralegals more efficient at “conducting research and preparing documents,” which may limit demand. It projects little or no change in paralegal and legal assistant jobs to 2035, against 5% growth for lawyers.

Our own scoring agrees on the direction. We scored typical task mixes with exactly the same method our free test uses on a real person’s week. A typical paralegal, whose week splits between writing, admin and research, scores 84, in our “very high” band next to bookkeepers and data entry clerks.

Not every measure agrees, which is worth knowing. Microsoft’s Copilot-based score for paralegals is low, in the bottom 15% of occupations, because few Copilot conversations matched the specific legal-support activities it assigns to them. The BLS, which combines five measures, rates paralegals “high.” The safe reading is that the research and drafting parts of the job are highly exposed, and the parts involving people, courts and judgment are not.

The job isn’t disappearing. The BLS still expects about 40,800 paralegal openings a year, mostly to replace people who retire or change careers. But the paralegals in the strongest position will be the ones who run the technology rather than compete with it. These three careers in our index show where legal support skills transfer well:

Compliance is the most natural move: every new rule about data and AI creates work for someone who understands regulation and can sign off. Setting up AI and automation for a legal team, or running legal projects and e-discovery, puts you on the side of the change that is growing.

Will AI replace lawyers in 10 years?

The official ten-year outlook says no. The BLS projections run from 2025 to 2035, and they show lawyer jobs growing 5% while legal secretary jobs fall about 5%. Nobody can promise what AI will do by 2035, but the BLS factored automation into these projections and still expects demand for lawyers to rise as businesses, governments and individuals need more legal help.

What will change within ten years is how junior lawyers learn. The tasks AI does best, such as research, document review and first drafts, are exactly the tasks that used to train first-year associates. Stanford’s Digital Economy Lab found that employment of workers aged 22 to 25 in AI-exposed occupations is now 19% below where it would be had it kept pace with less-exposed peers, while experienced workers show no comparable gap. Early-career lawyers should expect to be judged less on volume of research and more on judgment, client skills and how well they direct AI tools.

Will AI replace judges?

Very unlikely. Judges exercise public authority that the law gives to people, not software, and their work is about discretion, fairness and accountability. The BLS rates judges’ AI exposure as high, one level below lawyers, and projects their jobs to grow 2.8% by 2035. Microsoft’s data puts judges in the bottom quarter of occupations for AI applicability.

AI will still reach the courtroom. Judges already spend time catching AI-invented citations in the filings in front of them, and AI will likely help with research and court administration. But the decision, and the responsibility for it, stays with a person.

How to AI-proof a legal career

The same rule applies to lawyers, paralegals and legal assistants: let AI do the production and take ownership of the judgment. Here is what that means for each type of task our test measures.

For legal careers, five moves matter most:

1. Use AI for research and first drafts, and verify every citation. Lawyers who get reliable work out of AI will outpace those who refuse to use it. Lawyers who trust it blindly end up in the sanctions database.
2. Move toward advice, negotiation and advocacy. These are the least exposed parts of legal work, and they are what clients pay a premium for.
3. Own the client relationship. Explaining options, managing expectations and building trust can’t be delegated to software.
4. Specialize where the rules are new or messy. Areas such as AI regulation, privacy and complex disputes reward judgment over volume, which is where AI is weakest.
5. If you’re a paralegal, become the technology lead. Learn e-discovery and the firm’s AI tools, and be the person who checks their output. That makes you harder to replace, not easier.

For a full 90-day plan, see our guide to the skills AI can’t replace. The same pattern plays out in finance, as we show in Will AI replace accountants?, and our ranking of the 25 most AI-proof jobs shows where low exposure meets good pay.