TL;DR: The unauthorized practice of law (UPL) reserves legal advice - law applied to a specific person's situation - to licensed lawyers, while general legal information is broadly available to anyone. AI tools sit on both sides of that line. A licensed lawyer who uses AI as a tool and stays responsible is on safe ground; an unsupervised, direct-to-consumer tool that gives individualized legal advice raises real UPL questions that the law is still working out.
Artificial intelligence has made it trivial to generate something that looks like legal advice - a tailored answer to what should I do about my lease, delivered in confident, lawyerly prose. That capability runs straight into one of the oldest rules in the profession: only licensed lawyers may practise law. The unauthorized-practice-of-law doctrine exists to protect the public from unqualified advice, and it does not care whether the adviser is a person or a model. This article explains what UPL is, where the line falls, and why a lawyer-supervised tool and a consumer chatbot can end up on opposite sides of it.
What is the unauthorized practice of law?
The unauthorized practice of law is the provision of legal services - most importantly, legal advice - by someone who is not licensed to practise in the relevant jurisdiction. In the United States, each state defines and enforces UPL for itself, typically through its supreme court and bar, and the rationale is consistent everywhere: protecting the public from harm caused by unqualified or unaccountable advice. The remedies vary from injunctions to civil liability, and in some places UPL can carry criminal penalties.
Historically, UPL questions arose around non-lawyers who drafted documents, ran form-filling services, or counseled people on their rights for a fee. The doctrine is not aimed at ordinary speech - publishing a legal textbook or explaining the law on a website has long been fine - but at conduct that amounts to acting as someone's lawyer without a license. The American Bar Association maintains background on professional regulation at americanbar.org, and enforcement happens at the state level through bodies such as the State Bar of California. AI did not create these questions; it sharpened them by making advice-like output cheap and instantaneous.
Where is the line between legal information and legal advice?
The whole analysis turns on a single distinction: legal information versus legal advice. Legal information is a general statement of what the law is or how a process works, useful to anyone and tied to no one's particular situation. Legal advice applies the law to a specific person's specific facts and recommends a course of action, within a relationship where that person reasonably relies on it. Three features tend to mark the crossing point - individualization, application to specific facts, and reliance. The table makes the contrast concrete.
| Scenario | Legal information (generally permissible) | Legal advice (reserved to licensed lawyers) |
|---|---|---|
| A lease dispute | Most leases require notice before eviction, and notice generally means a written communication within a set period. | Based on your lease and these facts, you should withhold rent and send this specific letter today. |
| A contract term | An indemnity clause shifts certain risks from one party to another, and here is how such clauses typically operate. | In your agreement with this counterparty, accept the indemnity but cap it at the fees you have paid. |
| Court process | A small-claims case generally begins by filing a form with the court clerk and paying a fee. | Given your situation, file this particular claim for this amount before the deadline on Friday. |
| Estate planning | A will generally must be signed and witnessed to be valid, and intestacy rules apply if there is none. | For your estate and family, you should use a trust structured in this specific way. |
The line is real but not always crisp. The same sentence can be harmless information in a brochure and advice when delivered to one worried person about their exact problem. Context, the presence of a professional relationship, and whether the recipient is meant to rely on the statement all matter, which is why reasonable observers sometimes disagree about a given example.
How does UPL apply to AI tools?
Map that distinction onto software and the picture clarifies. An AI that outputs general legal information behaves much like a very capable legal encyclopedia or a self-help guide - longstanding categories that have not been treated as the practice of law. An AI that takes an individual's particular facts and tells them what they should do begins to resemble advice, and the harder it is to tell the output apart from what a lawyer would say, the sharper the UPL question becomes. The pivotal issues are whether the interaction is individualized, whether a licensed lawyer is responsible for it, and whether the user is invited to rely on it as advice.
This is not a brand-new debate. For years, direct-to-consumer legal document and self-help services faced UPL challenges in various states; many resolved those challenges by disclaiming the giving of legal advice, adding lawyer involvement, or otherwise staying on the information side of the line. Generative AI revives the same questions with more force, because the output is conversational, individualized on demand, and easy to mistake for counsel. Regulators have noticed, and bar guidance on AI - surveyed in our piece on what state bar AI ethics opinions say - increasingly touches these themes, even as the law remains unsettled as of 2026.
Why is lawyer-supervised AI different from direct-to-consumer AI?
The most useful way to think about UPL risk is to ask who is responsible for the output. That single question separates the two main models of legal AI, and it explains why one is comfortably established while the other sits in a gray zone.
Lawyer-supervised AI: the safe ground
When a licensed lawyer uses AI inside a representation, the lawyer is practising law and the AI is a tool - no different in principle from legal research databases, document assembly software, or the work of a paralegal. The lawyer supplies the judgment, verifies the output, and remains responsible for the final advice, much as the duty to supervise nonlawyer assistance already requires. There is no unauthorized practice because a licensed human is firmly in the loop and accountable for the result. That accountability is the whole point: it is why keeping a human in the loop is not just good practice but the thing that keeps AI-assisted work on the right side of the line - and why the malpractice exposure we discuss in AI malpractice risk for lawyers rests on the lawyer, not the software.
Direct-to-consumer legal AI: the gray zone
The harder case is a consumer-facing tool that answers individuals' legal questions with tailored recommendations and no lawyer involved. The more such a tool individualizes its output - taking a user's facts and telling them what to do - the closer it comes to the conduct UPL rules are designed to police. Vendors manage this risk in a few ways: by staying on the information side of the line, by displaying clear and prominent disclaimers that the tool is not a lawyer and does not provide legal advice, and by routing users to licensed lawyers for anything that requires judgment. Those measures help, but they may not fully cure conduct that is, in substance, individualized advice. As of 2026 the law here is genuinely unsettled, and reasonable regulators and commentators disagree about where exactly the boundary sits.
What does this mean for lawyers who use AI?
For practising lawyers, the reassuring takeaway is that using AI as a tool is squarely permissible and increasingly expected. You are not at risk of UPL for running research through an AI, generating a first draft, or reviewing documents at speed, provided you stay responsible - verifying the output, supervising it, and exercising your own judgment before anything is filed or advised. The flip side deserves attention too: do not let a tool effectively make the legal judgment for you in a way that delegates advice to the machine, and take care when supervising staff or guiding clients who may lean on consumer tools as a substitute for counsel. Treat AI output as a draft to be settled, not a conclusion to be forwarded. For the broader ethical frame, our overview of AI ethics in legal practice ties these duties together.
How is Judicio built for legal professionals?
Judicio is designed for the safe side of this line. It is a workspace built for legal professionals, not a direct-to-consumer advice service: it accelerates a lawyer's own research, review, and drafting rather than purporting to counsel a member of the public. One upload into the File Library feeds Legal Research, Document Review, the Review Matrix, Timeline Builder, Translation, and Drafting - and every research finding cites the exact page and quoted passage so the lawyer can verify it rather than take it on faith. Crucially, Judicio's outputs are explicitly not legal advice, and the platform is built on the assumption that a licensed professional keeps judgment and responsibility. That design keeps it firmly on the tool side of the information-advice divide: it makes a lawyer faster, but it does not replace the lawyer.
How do you get started?
If you are a legal professional, the way to see where Judicio sits is to use it on your own matters and watch how it supports - rather than supplants - your judgment. A 7-day free trial gives you 500 credits with no credit card required, so you can run research, review, drafting, timelines, and translation and confirm that the output is verifiable assistance you remain in control of. Professional plans are $200 per month for 5,000 credits. To discuss how the tool fits a professional, lawyer-supervised workflow, get in touch.
This article is general information about the unauthorized practice of law, not legal advice; AI outputs are not legal advice, and a licensed lawyer remains responsible for any legal judgment.