AI for Lawyers: What It Handles and What It Cannot
AI for lawyers: what it handles in research and drafting, where it fails, what the ABA ethics rules require, and how to protect client confidentiality.
Last Updated: September 2026
AI for lawyers means software that reads and writes language to help with legal work, such as research, drafting, document review and summaries. It is useful for the first pass on work that involves a lot of text. It is not reliable enough to work without a lawyer checking it, and it cannot take on a lawyer's professional duties.
This guide covers what AI handles well, where it fails, and what the ethics rules say about using it. It also covers the question every lawyer has to answer before using any tool: what happens to confidential client information.
Key Takeaways
- AI handles the first pass on research, drafting, review and summaries. The lawyer still owns the final work.
- AI legal research tools still make mistakes. A Stanford study found leading tools gave incorrect information in more than 17% of test queries.
- Courts treat AI errors as the lawyer's errors. In Mata v. Avianca, lawyers were penalized $5,000 for citing cases ChatGPT had invented.
- The ABA's first ethics opinion on AI, issued in July 2024, applies the existing duties of competence, confidentiality, communication and reasonable fees.
- Client information needs protecting before it goes into any AI tool, not after.
What Can AI Do for Lawyers?
AI is best at the first pass on work that is heavy on text and easy to check. These are the most common uses in legal practice:
- Summarizing depositions, case files and long contracts
- Drafting routine letters, memos and standard clauses
- Reviewing large document sets for relevant material
- Finding cases and statutes for a research question
- Comparing contract versions and flagging unusual terms
What these tasks share is that a person can check the output against a source. That is what makes them a good fit. The AI saves time on reading and first drafts, and the lawyer spends that time on review and judgment.
The time savings are real. The Thomson Reuters 2024 Future of Professionals report surveyed more than 2,200 professionals across legal, tax and risk. They expected AI to free up 12 hours a week within five years, and four hours a week in the first year.
The table below shows where AI fits across common legal tasks, and what stays with the lawyer.
How Reliable Is AI Legal Research?
AI legal research lets a lawyer ask a question in plain language instead of building keyword searches. The tool returns cases, statutes and a written answer. That can shorten the time it takes to find a starting point.
The weakness is accuracy. In a Stanford study of leading legal research tools, Lexis+ AI and Ask Practical Law AI gave incorrect information more than 17% of the time. Westlaw's AI-Assisted Research did so more than 34% of the time.
General chatbots did worse. The same team's earlier study found general-purpose chatbots were wrong on 58% to 82% of legal queries. So specialist tools are better, but none of them are reliable enough to trust without checking.
In practice, that makes AI research a starting point rather than an answer. Every case it cites needs to be pulled and read, and every summary needs checking against the source. The time saved on searching still counts, as long as the time spent checking is built in.
How Should Lawyers Use AI for Drafting?
Drafting is where many lawyers start, because the risk is easy to control. The AI produces a first version, and the lawyer decides what stays.
It works best on documents with a familiar structure. Examples include engagement letters, routine correspondence, first-draft clauses and file summaries for a colleague. The more standard the document, the less the lawyer has to change.
It works less well on anything that turns on the client's specific facts or on a live dispute. A settlement letter, for example, still needs the lawyer's judgment on tone, timing and what to concede.
A few habits make AI drafting safer:
- Give the tool the facts it needs, and leave out client details it does not need.
- Ask for a plain structure first, then refine the wording yourself.
- Check every legal statement, citation and date in the draft.
- Keep your own precedents and templates as the base where you have them.
The goal is a faster first draft, not a finished document. The final wording, and the responsibility for it, stay with the lawyer.
Where Does AI Fail in Legal Work?
AI fails in three predictable places: made-up information, missing context and judgment.
Made-up information
AI tools predict likely words rather than look up verified facts. That is why they can produce a case citation that looks real but does not exist. These errors are called hallucinations, and they are hard to spot because the tool states them with confidence.
The best-known example is Mata v. Avianca, a 2023 case in which lawyers filed a brief citing court decisions that ChatGPT had invented. The court imposed a $5,000 penalty on the lawyers and their firm. Our article on what happened when a lawyer relied on AI looks at the consequences in more detail.
Missing context
An AI tool only knows what it is given. It cannot know what the client said in a meeting, or how a judge tends to rule. Advice that ignores that context can be accurate on paper and wrong for the client.
Judgment
Strategy, negotiation and client counseling depend on weighing risk and reading people. Those choices carry professional responsibility, and they stay with the lawyer.
What Do the Ethics Rules Say About Lawyers Using AI?
In July 2024, the American Bar Association issued Formal Opinion 512, its first formal ethics opinion on generative AI. It does not create new rules. It explains how the existing Model Rules apply when lawyers use AI tools.
The opinion points to four duties in particular:
- Competence (Rule 1.1). Lawyers should understand the benefits and risks of the technology they use to serve clients.
- Confidentiality (Rule 1.6). Lawyers must keep information about the representation confidential unless the client gives informed consent.
- Communication (Rule 1.4). Lawyers must reasonably consult clients about the means used to reach their goals.
- Fees (Rule 1.5). Fees must be reasonable. A lawyer may charge for time spent entering information into an AI tool and reviewing the draft it produces. In most circumstances, a lawyer cannot charge a client for learning how to use the tool.
The ABA's Model Rules are a template, and each state adopts its own version. State and local bars are also issuing their own guidance on AI, so lawyers should check the rules where they practice.
How Should Lawyers Protect Client Confidentiality When Using AI?
Start by finding out what happens to anything typed into the tool. Terms differ between tools, and between free and paid plans of the same tool. Some providers store what users enter, and some use it to improve their models.
A few habits cover most of the risk:
- Read the tool's data terms before entering any client information: storage, training use and who can access it.
- Remove names and identifying details where the task does not need them.
- Get informed client consent where the confidentiality rules require it.
- Keep a record of which tool was used on which matter, and who reviewed the output.
Confidentiality is the question to settle first, because it cannot be fixed after the fact. Once client information has gone into a tool that keeps it, the lawyer no longer controls it.
What Does Good Human Review of AI Output Look Like?
Good review treats AI output like a first draft from a new junior colleague. It may be useful, but nothing leaves the office until someone accountable has checked it.
For research and filings, that means a few fixed steps:
- Pull and read every case the tool cites.
- Confirm every quote appears in the source, word for word.
- Check facts against the client file, not the tool's summary.
- Make sure the conclusion fits this client's situation, not just the general rule.
The lawyer who signs the work owns it. Courts have made that clear, and the ABA opinion builds on the same principle.
It also helps to keep a record of what was checked and by whom. If the work is questioned later, that record shows what was done. AI can shorten the path to a good draft, but it does not share the responsibility for it.
For a wider look at how AI is changing the profession itself, see our article on whether AI will replace lawyers.
Frequently Asked Questions
Is there AI for lawyers?
Yes. Lawyers use AI for research, drafting, document review and summaries. Specialist legal tools exist alongside general chatbots, and all of them need a lawyer to check the output.
Can lawyers use ChatGPT?
They can, with care. General chatbots were wrong on 58% to 82% of legal queries in research cited by Stanford. Lawyers need to verify every citation and protect client information before entering it.
Is AI legal research reliable?
Not on its own. A Stanford study found leading legal research tools gave incorrect information in more than 17% of test queries. AI research is a useful starting point, but every source needs to be read and confirmed.
Can AI give legal advice?
No. AI tools can help organize information and draft documents, but they cannot hold a license or take responsibility for advice. The lawyer gives the advice and answers for it.
What is an AI hallucination in legal work?
It is an answer that sounds confident but is false, such as a citation to a case that does not exist. It happens because AI tools predict likely words rather than check facts. In Mata v. Avianca, lawyers were penalized $5,000 for filing invented cases.
Do lawyers have to tell clients they use AI?
Sometimes. The ABA's Formal Opinion 512 points to the duty to reasonably consult clients about how their goals will be met. Informed consent may also be needed before client information goes into some tools, so lawyers should check their state's rules.
Can lawyers bill for time spent using AI?
Yes, within limits. Under the ABA opinion, a lawyer may charge for time spent entering information and reviewing the draft. In most circumstances, a lawyer cannot charge for time spent learning how to use the tool.
Does using AI put attorney-client confidentiality at risk?
It can. Entering client information into a tool that stores or reuses it can expose that information. The ABA opinion treats this under the duty of confidentiality, so lawyers should check a tool's data terms first.
Executive Summary
AI for lawyers is most useful on the first pass of text-heavy work: research, drafting, document review and summaries. It is not reliable enough to use unchecked.
Leading legal research tools gave incorrect information in more than 17% of Stanford's test queries, and courts hold lawyers responsible for AI errors. The ABA's Formal Opinion 512 applies the existing duties of competence, confidentiality, communication and reasonable fees to AI use.
The practical approach is to settle confidentiality first and use AI for drafts and starting points. An accountable lawyer then reviews everything that goes to a client or a court.
What Should You Do Next?
If you are a lawyer, pick one routine task, such as summarizing a deposition, and try an AI tool on it with a full review. That shows you both the time saved and the checking it needs.
If you run a firm, the useful question is where AI fits the work your team already does. Our page on AI for law firms covers how that is approached.
People Also Read
About the Author
Jerry Jariwalla is the founder of AiBuildrs and creator of the Growth Signal Intelligence framework. With over 22 years in digital marketing and multiple successful business exits, Jerry has spent the past decade leading AI implementation programs for mid-market businesses across professional services, recruitment, membership organizations, and traditional industries.
Expertise: AI Strategy, AI Implementation, Workflow Automation, Custom AI Development, Voice AI, Offshore Engineering, B2B Sales Intelligence, Mid-Market AI Adoption
Connect: LinkedIn
Disclaimer: This content is for informational purposes only and does not constitute professional business or technology advice. ROI outcomes vary based on industry, existing systems, and implementation commitment. Contact AiBuildrs for a consultation regarding your specific situation.
Jerry Jariwalla
Jerry Jariwalla, Founder & CEO of AiBuildrs, builds custom AI infrastructure, voice AI, automation systems, and The AEO Engine. View Jerry Jariwalla's Full Profile →