Legal Intake Compliance Guide: Disclosures, Rule 5.3 Supervision and the Unauthorized Practice Line
Intake is the first thing your firm says to a prospective client, so the professional-conduct rules apply from the first sentence. Three rules do most of the work. Say what the intake is and is not: software or a person, not a lawyer, no legal advice and no attorney-client relationship. Stop where collecting facts would turn into applying the law to them. And supervise it under Rule 5.3: know every sentence it says, and review what it collected. Recent bar opinions apply exactly these rules to website chatbots.
Why intake is where compliance is decided
Most of a firm’s compliance work happens after engagement: the letter, the conflict check, the file. Intake happens before any of that, often after hours and often without a lawyer in the room. It is also the first place a prospective client forms a view of what the firm promised them.
That is why the questions bar regulators keep asking about intake are the same three: does the person know who, or what, they are talking to? Did anyone tell them what their case is worth or what to do? And did a lawyer supervise what was said in the firm’s name? A front desk, an answering service, a contact form and an AI assistant all face those questions. The difference is how easy each one makes it to answer them.
This guide describes the ABA Model Rules and the opinions that apply them. Your own state’s version of each rule governs, and several states number and word them differently. Treat this as a map of the questions, not as an ethics opinion for your firm.
The three disclosures, in every conversation
Three statements carry most of the weight at first contact. They are different statements, and each one does a different job.
What the visitor is talking to. If intake is software, it should say so at once. The Florida Bar’s Ethics Opinion 24-1 (January 2024) says a generative AI chatbot that communicates with clients or third parties must comply with the advertising rules and include a disclaimer that it is an AI program, not a lawyer or an employee of the firm. Oregon’s Formal Opinion 2026-208 (February 2026) reaches the same point from Rule 7.1: a chatbot that does not identify itself as a chatbot rather than a person could be a misleading communication. Even the ABA’s general opinion on generative AI, Formal Opinion 512, flags Rule 7.1 as an issue and points to Florida’s chatbot analysis.
That nothing said is legal advice. The visitor should know, before they ask, that the conversation cannot tell them what to do. This protects them from relying on an answer, and it protects the firm from appearing to have given one.
That talking to intake does not make the firm their lawyer. Representation starts when the firm says it does, in a conversation with a lawyer or in writing. Florida’s opinion warns against a chatbot that is “overly welcoming”: one that gives legal advice, fails to identify itself as a chatbot at once, or lacks clear disclaimers that limit what the lawyer has taken on. A warm assistant is fine. One that sounds like it has already taken the case is not.
Three practical tests follow from this.
- Placement. The first two, what the visitor is talking to and that nothing is legal advice, belong in the first message of every conversation, not on a policy page the visitor never opens. The third should be ready as a fixed sentence for the moment the conversation turns to whether the firm is now the visitor’s lawyer.
- Wording. They should be fixed sentences that say the same thing every time. If the wording changes from one conversation to the next, nobody can supervise it, and nobody can show later what a given visitor was told.
- Language. A disclaimer has to be one the visitor can understand. A visitor writing in Spanish who receives the disclosures in English has not been told much. If the intake speaks a language, its disclosures should be in that language too. We look at what complete Spanish intake involves in our guide to bilingual answering services.
One question deserves a prepared answer: “Is this confidential?” The honest answer depends on the facts. What a prospective client tells a firm is protected under Rule 1.18, but a chat window is not a private meeting with a lawyer. A good intake answers plainly. It does not promise more privacy than it can deliver, and it asks for only basic contact details until the attorney takes over.
The unauthorized practice line: where intake stops
Software cannot be licensed to practice law, and neither can a receptionist or an answering-service operator. So the useful question is not whether intake may touch legal subjects. It has to, because the visitor came about a legal problem. The question is where collecting facts turns into applying the law to them.
There is no single national definition of the practice of law. The ABA’s task force drafted one in 2002 that was never adopted. It described the practice of law as applying legal principles and judgment to a person’s circumstances or objectives. Most state definitions turn on the same idea. Hearing a person’s facts is intake. Telling them what those facts mean for them is legal advice.
| Intake may | Intake must not | |
|---|---|---|
| What happened | Ask what kind of matter it is, roughly when, and where | Say whether the facts make a case |
| Money | Say the team will go over cost at the first meeting, or quote a fee statement the firm approved | Estimate what a claim is worth, or improvise a fee |
| Deadlines | Ask whether anything is urgent: a hearing date, a deadline, pressure to sign | Tell the visitor how long they have to file |
| What to do | Collect contact details and pass them to the attorney | Advise whether to accept an offer, move out, or talk to an insurer |
| Who handles it | Take the details and let the team decide | Tell the visitor the firm will or will not take the matter |
The hardest cases are the ones nobody asked about. A visitor mentions a custody hearing, and an assistant trying to be kind says that judges usually consider a child’s wishes. Nobody asked for legal advice, and the visitor just received some. The ban on advice applies to what the intake says, not only to what the visitor asks. Reassurance about the outcome (“you’ll be fine,” “that sounds like a strong case”) falls on the wrong side of the line for the same reason.
Oregon’s opinion makes the same point for chatbots: the lawyer must ensure the chatbot does not attempt to give legal advice, which could implicate the unauthorized practice of law. The practical test is simple. Ask your intake “Do I have a case?” and “What is it worth?” and read exactly what comes back. The right answer is a clear, polite refusal and a hand-off to the attorney, in words that do not sound like a brush-off.
Rule 5.3: you supervise what it says, not only who sells it
ABA Model Rule 5.3 requires lawyers who manage a firm to make reasonable efforts to ensure that the conduct of nonlawyers assisting the firm is compatible with the lawyers’ own professional obligations. The comment to the rule extends that to services outside the firm. An answering service is covered. So is a chatbot vendor.
The AI-specific opinions apply the same rule. ABA Formal Opinion 512 (July 2024) says managerial lawyers must establish clear policies on the firm’s use of generative AI, supervisors must see that lawyers and staff are trained, and the due diligence lawyers owe any outside vendor applies to AI providers too. Oregon’s opinion goes further for chatbots. The lawyer should understand how the tool works in enough detail to supervise it, do due diligence on the vendor, and be able to monitor and change what it says. It also notes that an autonomous AI agent, which can take actions without a person in between, carries more risk than a chatbot that only follows a defined path.
In practice, supervising intake comes down to four habits.
- Know every fixed sentence. You should be able to read, in advance, the exact words your intake uses for the disclosures, the refusals and any safety resources. If the vendor cannot show you the sentence, it is not fixed.
- Control changes. Ask who can change the wording or the questions, and how you will be told. A script that drifts is a script you are no longer supervising.
- Read what arrives. Review intake records and, where they exist, conversation transcripts promptly, and especially the first weeks after anything changes.
- Test it yourself. Run your own intake as a visitor would, after hours: a simple matter, an urgent one, and one outside your practice areas. Then read the record your firm received.
None of this requires a technical background. It requires the same thing supervising a new receptionist does: knowing what they say on your behalf, and checking.
What the first conversation collects, and what it protects
Rule 1.18 makes a person who consults a firm about possibly hiring it a prospective client, and what they share is protected even if the firm never takes the matter. Information that could significantly harm that person can disqualify the lawyer from representing someone on the other side, and the firm with them. We cover how that shapes the first conversation in our guide to answering services with legal intake. The short answer is to collect what the conflict check and the call back need, and leave the story for the attorney.
Oregon’s opinion adds two duties that apply squarely to website intake. First, when a chatbot gathers enough to identify a prospective client, the firm should enter that person in its conflict system promptly. That works only if the intake asks for the other party’s name and sends it to the firm in a form someone actually checks. Second, the opinion notes that a person in a distressed or impaired state may be recognizable to a human but not to a chatbot, so the lawyer should review those conversations promptly. An intake that recognizes a crisis, gives emergency and hotline resources at once, and flags the record as urgent makes that review possible. One that treats every visitor as a lead does not.
What the guidance says in the states where I practice
I am admitted in New York, Florida and Washington, D.C., and each has published generative-AI guidance. Florida’s Opinion 24-1 is the most specific about chatbots, as described above. The D.C. Bar’s Ethics Opinion 388 (April 2024) and the New York City Bar’s Formal Opinion 2024-5 (August 2024) build on the existing duties of competence, confidentiality and supervision rather than creating new ones. Most of the guidance across the country follows that pattern: no new rules, only the old ones applied to new tools.
If your state has not addressed chatbots directly, the Florida and Oregon opinions are a fair guide to the questions your bar would ask. If it has, its opinion governs. Either way, check your state bar’s ethics opinions before you rely on any intake tool, including ours.
An eight-point compliance check for any intake channel
Use this on a phone script, a contact form, an answering service or an AI assistant. It takes an evening, and most of it is reading.
- Does the first message say what the visitor is talking to? Software or a person, and not a lawyer, in the opening, not a footer.
- Are the disclosures fixed sentences you have read? Ask to see the exact wording, in writing.
- Are they in every language the intake speaks? Test one conversation in each.
- What does it say to “Do I have a case?” and “What is it worth?” Anything other than a polite refusal and a hand-off is a problem.
- Does it volunteer legal propositions nobody asked for? Describe a court date or an insurance call and read the reply.
- What does it say about cost? Only a fee statement your firm approved, word for word, or nothing beyond the first meeting.
- Does it collect the other party’s name before the story, and does that name reach your conflict check? Read a real record.
- Who can change the wording, and how will you know? Get the answer in writing.
The bottom line
Legal intake is a communication by the firm, so it carries the firm’s obligations from the first sentence. The three disclosures tell the visitor what they are dealing with. The unauthorized practice line keeps intake on the side of collecting facts. Rule 5.3 makes the lawyer responsible for knowing what is said in the firm’s name. None of these rules is new. What the recent opinions add is a clear statement that they apply to a chatbot exactly as they apply to a person. For how intake fits together as a whole, start with our complete guide to legal intake.
Common questions
Does a chatbot on a law firm website have to say it is AI?
The Florida Bar’s Ethics Opinion 24-1 says a chatbot that communicates with clients or third parties must include a disclaimer that it is an AI program and not a lawyer or employee of the firm. Oregon’s Formal Opinion 2026-208 says a chatbot that does not identify itself could be a misleading communication under Rule 7.1. Check your own state’s rules and opinions.
Can intake software tell a visitor whether they have a case?
No. Assessing whether someone’s facts make a case means applying legal judgment to their circumstances, which is the practice of law. Intake should decline politely and hand the question to the attorney.
Does a chat with a law firm’s intake assistant create an attorney-client relationship?
It should not, and the intake should say so. But the visitor may become a prospective client under Rule 1.18, which protects what they share and can create conflicts for the firm. That is why intake should collect only what the conflict check and the call back need.
Who is responsible when an intake chatbot says something wrong?
The lawyers. Under Rule 5.3, lawyers who manage a firm must make reasonable efforts to ensure that nonlawyer assistance, including outside vendors and their tools, is compatible with their professional obligations. ABA Formal Opinion 512 and recent state opinions apply that duty to AI tools.
Should a law firm chatbot promise that the conversation is confidential?
Not in absolute terms. Rule 1.18 protects what a prospective client shares, but a chat window is not a private consultation, and a promise of confidentiality may be more than the firm can deliver. A plain caution and a request for basic contact details only is safer.
Intake built to these lines. Lexara AI is an AI intake assistant built by an attorney. Its first message says it is an AI assistant, not an attorney, and that nothing in the conversation is legal advice, in English, Spanish or Haitian Creole, whichever the visitor writes in. The disclosures, the refusals and the safety resources are fixed sentences, so it says the same thing every time. It declines to evaluate cases or quote fees your firm has not approved, collects the other party’s name for your conflict check, and sends nothing until the visitor confirms the details. Try the live demo on our homepage.
Related: Legal Intake, Explained — our complete guide · Answering service with legal intake: how to choose · Legal client intake software: a small-firm buyer’s guide · Bilingual answering service for law firms · Legal intake software: the 7 features that matter · all guides.