Florida Bar AI chatbot rules: a build checklist for personal injury intake

The Florida Bar AI chatbot rules that matter most for intake come from Florida Bar Ethics Opinion 24-1: an AI intake chatbot or receptionist should tell prospective clients they are talking to an AI program, identify itself as a nonlawyer, stick to factual questions, give no legal advice, route legal questions to a lawyer and screen for people who already have one. The firm stays responsible for everything it says. Florida’s solicitation rule and all-party recording consent law add two more constraints: answer rather than cold-call, and announce the recording.

This is a build checklist from the team that builds these systems, not legal advice. Before launch, confirm your scripts and settings with The Florida Bar’s Ethics Hotline (a service for Bar members) and your own ethics counsel. For the broader picture, see our page on AI intake for Florida personal injury firms.

What are the Florida Bar AI chatbot rules?

There is no single Florida rule titled “AI chatbots.” The guidance comes from Florida Bar Ethics Opinion 24-1 on lawyers’ use of generative AI, which the Bar’s Board of Governors approved on January 19, 2024. LawNext’s coverage notes the vote was unanimous and the opinion is non-binding. Advisory or not, it shows how the Bar reads its existing rules when a lawyer puts AI in front of the public.

The short version: lawyers may use generative AI, with caveats in four areas. Hinshaw & Culbertson’s summary lists them as confidentiality, lawyer oversight, fees and costs, and advertising. For an AI receptionist, five sets of obligations shape the build:

• The opinion’s chatbot guidance: disclosure, nonlawyer status, facts only, no legal advice, screening for existing representation.
• Advertising rules, including Rule 4-7.13 on deceptive and inherently misleading advertisements. What your bot says about the firm counts.
• Supervision of nonlawyer assistance, which the opinion applies to AI, including AI run by a third party.
• Confidentiality, which turns on how your AI provider stores, shares and learns from conversations.
• Solicitation and recording: Rule 4-7.18 limits who the agent may contact; Florida’s recording statute governs how it records calls.

What does Florida Bar Opinion 24-1 say about AI intake chatbots?

The opinion speaks directly to chatbots that talk with prospective clients. Each point becomes something concrete in the build.

1. Say it is AI, in the first sentence

Per the Hinshaw summary, a lawyer must inform prospective clients that they are communicating with an AI program. LawNext flags the risk of a chatbot that does not immediately identify itself as one. On the phone, that means the greeting; in chat and DMs, the first message.

2. Make clear it is not a lawyer

“AI assistant” alone is not enough. The opinion says an intake chatbot should clearly identify its nonlawyer status. Someone calling from an emergency room will not infer that from a friendly voice.

3. Ask for facts, and only facts

The guidance is to limit questions to obtaining factual information: what happened, when, where, who was hurt, what treatment, who else was involved. Nothing in the script should invite the agent to interpret what the facts mean.

4. Refer every legal question to a lawyer

“Do I have a case?” “What’s it worth?” “Should I give the adjuster a recorded statement?” The agent answers none of them and offers a path to a lawyer instead: a live transfer, a booked consult or a same-day callback.

5. Screen for existing representation early

The opinion suggests considering screening questions that limit the chatbot’s communications if another lawyer already represents the person. Ask “Are you already working with a lawyer on this?” early. If yes, the agent stops collecting details and closes politely.

6. Remember the firm owns what the bot says

The lawyer is ultimately responsible if the chatbot gives misleading information or communicates in a way that is inappropriately intrusive or coercive. Overconfident reassurance (“you definitely have a case”), claims about results and pushy follow-up all land on the firm.

A Florida Bar AI chatbot rules checklist for your build

Here is how each requirement turns into agent behavior and a way to check it. Use it as the agenda when your supervising attorney signs off on the system.

Requirement	Source	What the AI agent does	How you check it
Disclose it’s AI	Opinion 24-1	Fixed opening line on every channel, before any question	Test call and DM on every channel before launch
Nonlawyer status	Opinion 24-1	Says “not a lawyer” in the greeting and whenever asked	“Are you a lawyer?” in every test run
Facts only	Opinion 24-1	Attorney-approved question list: incident, date, location, injury, treatment, other parties, contact	Attorney signs off on the list
No legal advice	Opinion 24-1	Deflection script for case value, fault, deadlines and insurers, then a handoff to a lawyer	20+ legal test questions, rerun after every change
Already represented?	Opinion 24-1	Early screening question; “yes” ends intake politely	Weekly review of represented closes
Not misleading	Rule 4-7.13; Opinion 24-1	Only approved statements about the firm: no outcome predictions, no past results	Attorney approves every script change
Not intrusive or coercive	Opinion 24-1	Capped follow-ups; stops on “no” or “stop”	Monthly follow-up log review
Inbound only	Rule 4-7.18	Answers calls and messages the person started	Outbound dialing disabled in the platform
Recording consent	Fla. Stat. § 934.03	Recording notice at the top of every call; chat says conversations are saved	Weekly sample of call openings
Confidentiality	Opinion 24-1	Contracted confidentiality, breach notice and retention; no training on client conversations; role-based access	Written vendor answers on file
Supervision	Opinion 24-1	Intake manager reviews transcripts weekly; written escalation rules	Dated review notes and escalation log

Sample opening lines for a Florida law firm AI receptionist

First, hard-code the disclosure. The greeting should be a fixed line the system plays or sends word for word, because a model that paraphrases can drop “not a lawyer” on the one call that matters.

Phone (voice agent)

“Thanks for calling [Firm] Law. I’m the firm’s AI intake assistant, not a lawyer, so I can’t give legal advice, but I can get your information to our attorneys right away. This call is recorded. Are you calling about a new accident or injury, or are you a current client?”

Website chat

“Hi, I’m [Firm]’s AI intake assistant. I’m not a lawyer and can’t give legal advice, but I can take down what happened and set up a time for you to speak with an attorney. This conversation is saved and reviewed by our team. Can you tell me a little about what happened?”

Instagram, Facebook and TikTok DMs

“Thanks for reaching out to [Firm]. Quick note: this is our AI assistant, not a lawyer, so I can’t give legal advice, but I can get you to one fast. First, are you already working with a lawyer on this?”

When someone asks a legal question

“That’s a question for one of our attorneys, and I don’t want to guess. I can connect you with someone now, or book a time today. Which would you prefer?”

After hours, swap the live transfer for a same-day callback. Never let the agent promise a transfer it cannot make.

When someone is already represented

“Thanks for letting me know. Since you already have a lawyer on this, they’re the best person to talk to about it. I’ll stop here.”

If you want a line for people seeking a second opinion, decide that policy with counsel and script it word for word. For a full call script covering existing clients, conflict checks and booking, see what an AI receptionist for a law firm should handle.

Can a Florida law firm’s AI voice agent make outbound calls to accident victims?

Design it to answer, not to dial. Rule 4-7.18 prohibits soliciting prospective clients in person or by telephone, including real-time electronic contact, unless there is a family or prior professional relationship. Written solicitations about a personal injury or wrongful death matter cannot go out until 30 days after the incident. The Bar actively watches for violations after disasters, as its notice after Hurricane Ian shows.

A voice agent that could dial hundreds of numbers an hour makes this a configuration question, not just a policy one. In practice:

• Inbound only. The AI voice agent answers calls. Outbound dialing to numbers that never contacted the firm is switched off in the platform, not merely discouraged in the prompt.
• Follow-up only to people who reached out. Missed-call texts, confirmations, reminders and requested callbacks go only to people who called, messaged or filled in a form first.
• Reply in threads the person started. AI chat agents answer DMs and web chats; they do not message people who merely liked or commented on a post. Ask counsel about public comments; the conservative default is to wait for a message.
• No purchased or scraped lists. Lead lists, crash-report data and accident alerts never feed the agent.
• A hard stop on every sequence. “Stop,” “no thanks” or a set number of unanswered attempts ends it, which also addresses the opinion’s concern about intrusive communication.

How should AI intake at a Florida law firm handle recording and confidentiality?

Recording: Florida is an all-party consent state

Under Fla. Stat. § 934.03, as the Reporters Committee’s Florida guide explains, all parties to a confidential communication must consent to its recording, and violations can carry criminal and civil penalties. That is the reason most firms put a recording notice at the top of the call, before the caller says anything substantive.

Plan for the caller who objects: the agent offers a callback from a staff member, and the platform stops recording that call. Your counsel can tell you how the notice should read and whether it covers a transfer to an attorney.

Confidentiality: questions for your AI provider

Opinion 24-1 expects lawyers to understand how a third-party AI tool keeps, shares and learns from what it is told. Per Hinshaw, it also recommends informed consent before using a third-party tool that would disclose confidential information; ask your ethics counsel how that applies at intake. Get written answers to these:

1. Where are call recordings, transcripts and intake records stored, and can we set how long they are kept?
2. Is any conversation used to train or improve models that serve other customers? Put the answer in the contract.
3. Is the provider contractually bound to confidentiality, and how fast will it notify us of a breach?
4. Which of the provider’s staff can open transcripts, and is that access logged?
5. Can we export everything and have it deleted if we leave?

Minimize data, too. The agent should not ask for Social Security numbers, policy numbers or medical records in a chat window; evidence goes to a secure upload link tied to the lead. On our builds, the firm owns the code, prompts and data.

How should a Florida firm supervise its AI receptionist?

Per Hinshaw, the supervision rule requires policies that keep nonlawyer conduct, AI included, consistent with the lawyer’s obligations, and a third-party AI does not remove that duty. In operational terms:

• Name an owner. An intake manager runs the system day to day; an attorney supervises it and signs off on changes.
• Review transcripts on a schedule. Daily for two weeks, then weekly: a random sample, every escalation, every decline and every conversation with a legal question.
• Write escalation rules down. Catastrophic injury or death, distress, a legal question asked twice, a request for a person and anything the rules do not cover go to a human.
• Control changes. Edits to the greeting, disclosures, questions or statements about the firm need attorney approval and a dated change-log entry.
• Rerun the test set (case-value question, already represented, refuses recording, distressed caller, Spanish speaker) after every change.
• Keep an off switch that routes everything to staff or voicemail in one step.

What this looks like in a real Florida PI build

Darfoor Law Firm is a Fort Lauderdale, Florida personal injury practice. We built its AI intake in four weeks: chat agents on two Instagram accounts, Facebook Messenger and TikTok, plus a voice AI receptionist, all running the same logic. Two parts of that build line up with the opinion’s guidance:

• Representation is a hard gate. The five gates are Florida residency, confirmed physical injury, unrepresented status, a valid phone number and a complete name.
• A person takes the urgent calls. The voice agent briefs Mr. Darfoor on the situation, then connects the call, so he starts with full context.

Follow-up stays inbound: a missed call triggers two texts within seconds, one acknowledging the person who called and one alerting staff. The recorded results in the Darfoor Law Firm case study:

• 4 → 1: Instagram (two accounts), Facebook Messenger, TikTok and phone unified into one intake pipeline.
• 100% of inbound leads captured and answered, regardless of time of day or staff availability.
• 5 qualification gates applied automatically before anything reaches an attorney.

Where to start

1. List every entry point: each phone number, the website chat and every social account. Each needs its own fixed disclosure line.
2. Draft the five scripts above and have your supervising attorney mark them up.
3. Send your AI provider the confidentiality questions and file the written answers.
4. Take the plan to The Florida Bar’s Ethics Hotline and your counsel before anything goes live.
5. Launch after hours first, review transcripts daily, and widen coverage once reviews come back clean.

For the qualification questions, escalation triggers and metrics behind a PI build, read our guide to AI intake for personal injury law firms, or start with the free Case-Flow Checkup for PI firms.