Key takeaways
- Decide who answers chat (staff, a vendor or a supervised AI tool) and when, and say so plainly in the chat window.
- Florida's guidance says AI chatbots must disclose they are AI and not lawyers[1]; human chat agents should say they are not attorneys.
- Half of US adults are more concerned than excited about AI in daily life[2], so be open about who is on the other end.
- Collect a name and number first, conflict names before the story, and hand off to a call quickly.
What chat is good for, and what it is not
Website chat suits people who are reading about your firm, very often on a phone (91% of US adults own a smartphone[3]), and have a question they would rather type than say aloud: whether you handle their kind of matter, whether you have a Spanish-speaking attorney, how consultations work. It is quick, private and works on a phone. For many visitors it is the lowest-effort way to make first contact.
Chat is poor at some of the things intake needs most. Long stories get typed in fragments, urgency is easy to miss, and the conversation can drift into legal questions faster than on the phone, because typing feels informal. The goal of chat intake is therefore narrow: capture contact details, check fit, spot urgency and move the person to a consultation or a call.
Choose who answers
Firms have three realistic options, and many combine them by time of day.
- In-house staff: best quality and consistency with your phone intake, but only during the hours someone is free to watch the chat window.
- A chat vendor with trained agents: wider hours and fast replies, but the agents must use your script, your acceptance criteria and your handoff rules.
- A supervised AI chat tool: always available, but it must be configured to collect facts and hand off, never to answer legal questions, and the firm must supervise it.
Whichever you choose, put the hours and the type of responder in writing, and show them in the chat window. If chat is staffed only from 9 to 5, say so after hours and offer a callback request instead of leaving people typing into a void.
Be open about who is answering
Visitors want to know whether they are talking to a person or a program. Pew's 2025 survey found 50% of US adults are more concerned than excited about the increased use of AI in daily life, 38% are equally concerned and excited, and 10% are more excited than concerned[2]. And 76% say it is extremely or very important to be able to tell whether text was made by AI or by people[2].
Professional rules point the same way. The Florida Bar's 2024 opinion says generative AI chatbots that communicate with clients or third parties must include a disclaimer that the chatbot is an AI program and not a lawyer or employee of the firm[1]. Human chat agents should, in the same spirit, say they are not attorneys. California's State Bar guidance adds that lawyer supervision of an AI intake system must correspond to its level of access and autonomy[4].
The opening message
The first message sets the boundaries. Keep it short enough to read on a phone, and make it true for whoever is answering.
Human agent: "Hi, I'm [first name] from [firm]. I'm not an attorney, but I can help you get to the right lawyer here. What's your name, and the best number to reach you if we get disconnected?"
AI tool: "Hi, I'm [firm]'s virtual assistant. I'm an AI program, not a lawyer or a member of staff, and I can't give legal advice. I can take some details and help you book a consultation. What's your name, and the best number to reach you?"
Underneath, show a one-line notice that chatting does not create an attorney-client relationship and that visitors should not share confidential details until the firm has checked for conflicts. The Florida Bar warns that intake chatbots risk creating a prospective-client or even a lawyer-client relationship without the lawyer's knowledge[1], which is exactly why the notice and the question order matter.
The chat question order
Chat must ask fewer questions than the phone script, because every message is effort for the visitor. Use this order:
- Name and phone number. If the chat drops, you can still call.
- Matter type, offered as quick-reply buttons.
- State and county.
- Any court date, deadline or custody issue. If yes, prompt the visitor to call now and alert the on-call person.
- Whether they already have a lawyer on this matter. Florida's guidance suggests screening questions that limit a chatbot's communications with people who are already represented[1].
- Names of the other people or businesses involved.
- Offer two consultation times, or a call within a set time.
That is enough to book. Anything more belongs on the call or in the post-booking questionnaire.
Keep chat out of legal advice
Chat visitors ask legal questions constantly, and the informality of typing makes it tempting to reply with something quick. Give every chat agent, and configure every chat tool with, the same redirect pattern: acknowledge, explain who can answer, offer a time.
- Visitor: "Can my employer fire me for this?" Agent: "That's exactly the kind of question an attorney can answer once they know the details. I can book you a consultation. Would tomorrow at 10 or 2 work?"
- Visitor: "What are my chances?" Agent: "I'm not able to say, but the attorney can go through that with you. Can I take a few details so they're ready for your call?"
- Visitor pastes a long account of the facts: "Thanks for explaining. So the attorney can look at this properly, could you tell me the names of the other people involved first? We check those before going into detail."
Florida's Ethics Opinion 88-6, which the Florida Bar calls especially useful for intake chatbots, says nonlawyers doing intake should limit questions to facts, give no legal advice and refer legal questions to the lawyer[1].
Hand off fast
Escalate urgent chats live
If the visitor mentions custody, a hearing within your urgent window or a safety concern, the agent offers an immediate call and alerts the on-call attorney. The chat window shows the firm's phone number in large text.
Move to a call for anything complex
After the core questions, offer a call: "It might be quicker to talk. Can I have someone call you in the next 15 minutes?"
Write the chat to the intake record
The transcript and the structured answers go to the same intake record as phone and form leads, not to a separate chat dashboard nobody checks.
Close with a confirmation
Booking details go by email and, with permission, by text, as they would from a phone booking.
Follow up abandoned chats
If the visitor left after giving a phone number but before booking, call within your speed-to-lead target.
Example: a chat that turned into a booked call
Example (fictional): at 7:40 p.m. a visitor on an immigration firm's website opens chat and asks, "Can I still apply if my visa expired last year?" The firm's supervised AI chat tool says it is an AI program, not a lawyer, and says it cannot answer legal questions but can help book a consultation. It asks for her name and phone, the type of matter, and whether she has any hearing scheduled. She says no hearing, but she has an appointment notice for biometrics next week.
The tool records the appointment date, asks whether she already has a lawyer (no), takes her preferred language (Spanish) and offers two consultation times with a Spanish-speaking attorney. She books the 9 a.m. slot. The intake record the attorney reads in the morning has the same fields as a phone intake, with the appointment date at the top.
Measure chat honestly
- Time to first reply, by hour of day.
- Share of chats that capture a phone number.
- Share of chats that end in a booking or a scheduled call.
- Chats where the agent or tool answered a legal question, found in a monthly transcript review.
- Urgent chats, and how fast each reached an attorney.
Frequently asked questions
Should a law firm have live chat on its website?
It can help, especially for people who prefer to type or cannot talk privately, if someone answers quickly and the chat collects contact details, checks fit and hands off to a call or consultation.
Do law firm chatbots need a disclaimer?
Florida's Bar says AI chatbots must disclose they are an AI program and not a lawyer or firm employee[1]. Check your state's guidance and disclose either way.
Can a chat agent answer legal questions?
No. Chat agents and tools should collect facts and redirect legal questions to an attorney[1].
Can chat create an attorney-client relationship?
It can create prospective-client duties without the lawyer realizing[1], which is why chats should collect conflict names before details and show a clear notice.
References
The Florida Bar (2024). Florida Bar Ethics Opinion 24-1 (opens in a new tab) โ
Pew Research Center (2025). How Americans View AI and Its Impact on Human Abilities, Society (opens in a new tab) โ
Pew Research Center (2026). Americans' Use of Mobile Devices in 2026 (Mobile Fact Sheet) (opens in a new tab) โ
The State Bar of California, Standing Committee on Professional Responsibility and Conduct (2026). Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law (opens in a new tab) โ