Key takeaways
- Florida requires AI chatbots that talk to clients or third parties to say they are an AI program, not a lawyer or firm employee[1].
- Website intake chatbots can create prospective-client relationships without a lawyer knowing, so limit what they collect[2].
- Half of US adults are more concerned than excited about AI in daily life, so plain disclosure and an easy route to a person matter[3].
- Set it up like a new staff member: written scope, approved wording, testing and regular transcript review.
What a website chatbot should do for a law firm
A law firm chatbot has a narrow job: welcome visitors, answer basic questions about the firm (practice areas, office locations, how consultations work), take the details the firm needs to follow up, and hand serious conversations to a person. It is not a lawyer, and it should never try to sound like one. The firms that get value from chat treat it as an extension of intake, not as a legal information service.
That narrow job still carries professional duties. A chatbot speaks for the firm, gathers information from people who may become clients, and can create expectations the firm has to meet. Most of the setup work is about drawing those boundaries clearly.
The clearest guidance so far
Florida Bar Ethics Opinion 24-1 is the most detailed state guidance on intake chatbots, and many firms outside Florida use it as a reference. It makes four points that translate directly into setup decisions:
- Disclose. 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].
- Watch for relationships. Website intake chatbots risk creating a prospective-client or even lawyer-client relationship without the lawyer's knowledge[2].
- Follow the nonlawyer intake rules. The opinion points to Ethics Opinion 88-6: identify nonlawyer status, limit questions to factual information, give no legal advice and refer legal questions to the lawyer[4].
- Screen for representation. Consider screening questions that limit the chatbot's communications if the person is already represented by another lawyer[5].
Check your own state's ethics opinions and advertising rules as well, since chat messages are communications about the firm's services.
Visitors arrive wary
Public attitudes shape how a chatbot should behave. Pew found 50% of US adults are more concerned than excited about the increased use of AI in daily life, 38% are equally excited and concerned, and only 10% are more excited than concerned[3][6]. For a law firm, that argues for a chatbot that announces itself plainly, keeps conversations short, and makes it easy to reach a person or leave details for a call back.
- More concerned than excited 50%
- Equally excited and concerned 38%
- More excited than concerned 10%
Source: Pew Research Center, 2025[3]
Show the data
| Item | Value |
|---|---|
| More concerned than excited | 50% |
| Equally excited and concerned | 38% |
| More excited than concerned | 10% |
Set it up in eight steps
Define the scope
Write what the chatbot may do (firm information, intake details, booking or callback requests) and what it may not (legal advice, case evaluation, fee quotes beyond what the firm publishes).
Write the disclaimer
Have a lawyer approve the opening disclosure and a shorter reminder for later in the conversation.
Build the screening questions
Ask early whether the visitor already has a lawyer for the matter and which practice area applies.
Limit the data
Collect contact details, practice area, names for a conflict check and a short description. Leave detailed facts for a person.
Set the hand-off
Decide when the chatbot offers a person, a callback or a booking, and who receives each request.
Configure retention and access
Decide how long transcripts are kept, who can read them, and whether the vendor may use them for anything.
Test it hard
Try to make it give advice, promise representation or discuss another client. Fix every failure before launch.
Launch and review
Read every transcript for the first two weeks, then a regular sample.
What to collect and what to leave for later
| Item | Collect in chat? | Why |
|---|---|---|
| Name, phone, email | Yes | Needed for follow-up |
| Practice area | Yes, from a list | Routes the inquiry |
| Already has a lawyer? | Yes, early | Limits the conversation if represented |
| Names of other parties | Yes | Needed for the conflict check |
| Short description | Yes, a sentence or two | Lets staff prioritize |
| Detailed facts and documents | No | Wait until conflicts are cleared and the firm decides it can help |
| Health, financial or immigration details | No | Sensitive; collect later through secure channels |
| Questions about the case's strength | No answer given | Legal questions go to a lawyer |
Conversation design that keeps you safe
- Open with the disclosure, then one question at a time.
- Use buttons for practice areas and yes or no answers, which reduces misunderstandings.
- When a visitor asks a legal question, say it is a question for a lawyer and offer a callback or booking.
- Never use phrases like "you have a strong case" or "we can win this", even as encouragement.
- End every conversation with what happens next and when, without promising a lawyer's availability.
- Offer a non-chat route, such as a phone number, on every screen.
Sample wording
Sample opening: "Hi, I'm the firm's AI assistant. I'm not a lawyer or a member of staff, and I can't give legal advice. I can answer questions about the firm and take your details so the team can follow up."
Sample screen: "Before we continue, do you already have a lawyer helping you with this matter?"
Sample legal-question reply: "That's something one of our lawyers would need to look at. Would you like to request a call back or a consultation?"
Hand-off to a person
The hand-off is where many chatbots fail. Decide which conversations go straight to a person during office hours, which become callback requests, and which can book a consultation directly. Send each request to a named role, not a shared inbox nobody owns, and include the transcript so the visitor does not have to repeat everything. If a visitor mentions danger or an emergency, the chatbot should direct them to emergency services immediately and flag the conversation for review.
Design for phones
Many visitors will use chat on a phone, and 16% of US adults are smartphone-dependent, with a smartphone but no home broadband[7]. Check that the chat window does not cover the page's phone number, that the disclosure is visible without scrolling, that buttons are large enough to tap, and that visitors can close the window easily. A chatbot that is hard to dismiss on a small screen drives people away from the site entirely.
Advertising and marketing rules
Chat messages are communications about the firm's services, so your state's advertising rules apply to them as they do to your website. Avoid statements about results, comparisons with other firms and claims about expertise that your rules restrict. Do not let the chatbot quote past case results unless they appear on your site with the disclaimers your state requires, and never let it generate testimonials or reviews.
Review transcripts on a schedule
- Weeks one and two: read every transcript.
- After that: read a set sample each week, plus every conversation flagged for a legal question, an emergency or a complaint.
- Log any reply that went outside scope and fix the configuration the same week.
- Check that callbacks requested through chat were made, and how quickly.
- Delete transcripts on the retention schedule you set, and confirm the vendor does the same.
Example: a personal injury firm's chatbot
Example (fictional): an injury firm launches a chatbot limited to firm information and intake details. In the first week's transcripts, a visitor asks "Is my case worth suing over?" and the chatbot replies with a general description of negligence. The firm treats it as an out-of-scope answer, changes the configuration so any question about a case's strength gets the approved legal-question reply, and re-tests. A second visitor mentions an existing lawyer; the screen works, and the chatbot suggests they speak with that lawyer unless they want to discuss changing firms.
Questions to ask a chatbot vendor
- Can we set the exact opening disclosure and keep it visible throughout the chat?
- Can we limit answers to firm information we supply, so it never answers general legal questions?
- How are represented-person and emergency answers configured and tested?
- Where are transcripts stored, for how long, and are they used to train anything?
- How are requests handed to our staff, and what happens if nobody picks one up?
- Can we export every transcript if we leave?
Ask for the answers in writing, and test each one yourself during the trial rather than relying on the demo.
Common mistakes
- A disclaimer buried in terms of use instead of shown in the chat.
- An avatar photo that suggests a real person.
- Asking for the whole story before checking for conflicts or existing representation.
- No owner for chat requests, so callbacks are missed.
- Never reading the transcripts after launch.
Frequently asked questions
Can a law firm use an AI chatbot on its website?
Yes, with safeguards. Florida guidance requires a disclaimer that the chatbot is an AI program and not a lawyer or employee, and warns about accidentally creating client relationships[1][2].
What should a law firm chatbot never do?
Give legal advice, evaluate a case, promise representation or collect detailed facts before conflicts are cleared.
Should the chatbot ask whether the visitor has a lawyer?
Yes. Florida guidance suggests screening questions that limit communications with people already represented[5].
How often should chat transcripts be reviewed?
Every transcript for the first two weeks, then a regular sample plus every flagged conversation.
References
The Florida Bar (2024). Florida Bar Ethics Opinion 24-1 (opens in a new tab) โ
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
The Florida Bar (2024). Florida Bar Ethics Opinion 24-1 (opens in a new tab) โ
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