Quiet office workspace with laptops and headsets on the desks

An AI chatbot on a law firm website should do what a good receptionist does: answer routine questions, collect contact and matter details, screen and book consultations, then hand the conversation to a person. It works when its script is narrow, it is tied to your calendar and intake process, and someone at the firm reads its transcripts on a schedule. This guide covers what a firm chatbot should and should not do, which safeguards bar guidance points to, and how to judge a setup before you sign anything.

It is written for owners of solo and small firms who want more website visitors to become booked consultations without adding front-desk headcount.

The short version

  • The American Bar Association names FAQs, intake basics, prospective-client triage and scheduling as the practical jobs for a firm chatbot, in its guide to hiring a chatbot.
  • Treat the bot like front-line intake staff: write down what it may say, what it may never say, and when it must pass the visitor to a person.
  • Disclose that visitors are talking to an AI, build hard stops for complex or sensitive inquiries, and test conversations before launch and regularly afterward.
  • Judge any setup on data handling, calendar and CRM connections, and how the bot behaves when it does not know the answer, not on how fluent it sounds.

A firm chatbot collects and routes; it never tells a visitor they have a case

A law firm chatbot is not a legal assistant for the public, and it is not a research tool for attorneys. It is an intake tool that lives on your website, and in some setups on your Google Business Profile or text channels. Its job is to meet a visitor at the moment of interest and move them toward a conversation with your firm.

That distinction matters because search results for this topic mix three different products: consumer-facing legal Q&A bots, research assistants for attorneys, and intake chatbots for a firm's own website. This article is about the third. The ABA's guide frames the right mental model: delegate chatbot authority the way you would delegate to front-line intake staff, with guardrails. A receptionist collects information, schedules and routes. A receptionist does not tell a caller they have a case.

Handles

Office hours, practice areas, location, how a consultation works, what to bring, and booking a time that matches your calendar.

Collects

Name, contact preference, practice area, a short description, and the screening answers your staff need to prioritize.

Never does

Evaluates a visitor's situation, predicts an outcome or promises success. Anything that sounds like an assessment goes to a person.

The ABA guide lists FAQs, appointment setting, reminders and some parts of intake as the practical starting points. The same guide warns that a careless deployment can produce inappropriate conversations, misleading legal conclusions and guarantees of success. Your configuration should make those impossible, not merely unlikely.

The after-hours visitor who finds only a contact form is the gap a chatbot closes

Most small firms already pay to get a visitor to the site, through local SEO, Google Ads or Local Service Ads. The weak point is often what happens next. Someone with an urgent problem lands on a page at night, finds a contact form, and moves on to the next firm in the results. A chatbot answers right away and offers a consultation slot. Our guide to law firm intake response time covers why that first reply matters so much.

The ABA guide notes that chatbots work around the clock and give visitors a sense of instant responsiveness. Whether that turns into more consultations for your firm depends on your traffic, your practice area and how fast your team follows up. What you control is how well the bot is set up and how quickly someone responds to the leads it hands over. If you want to know what share of your visitors become inquiries today, start with our explainer on law firm website conversion rate so you have a baseline before you change anything.

A seven-step script keeps the bot short, disclosed and easy to audit

A well-built flow is short, respectful and predictable. This sequence is a sensible default for most small firms.

  1. Greeting with disclosure. The bot says it is an automated assistant for the firm and that a member of the team will follow up.
  2. Scope check. It asks what kind of matter the visitor needs help with and compares the answer with your practice areas and geography.
  3. Core details. Name, phone or email, preferred contact method and a brief description. It does not ask for sensitive detail it does not need.
  4. Screening prompts. Questions you define, such as whether the person already has an attorney on the matter or faces an urgent deadline.
  5. Routing. Qualified inquiries go to staff with a summary of the conversation. Urgent or complex ones are flagged for immediate attention.
  6. Scheduling. The bot offers consultation times that follow your firm's rules and sends a confirmation.
  7. Respectful exit. Matters outside your practice get a polite message and, where appropriate, a suggestion to look for a different resource, without any comment on the merits.

The screening in step 4 is not cosmetic. The ABA's article on AI in firm marketing says prospective-client screening should guard against improper disclosure and identify people who are already represented.

Bar guidance lines up on disclosure, hard stops and human review

Rules differ by state, so check your own state bar's guidance. The published sources below agree on a common core, and a chatbot configured around it is easier to defend and easier for visitors to trust.

Say it is an AI, cap what it can discuss, and hand off early

The ABA's September 2024 article recommends disclosing chatbot use, handing intake to a professional at the proper time, building hard stops for complex inquiries and testing conversations. It also reports that Florida's advisory Opinion 24-1 makes lawyers responsible for chatbot communications that are misleading or inappropriately intrusive or coercive. Wherever your firm practices, that is the practical reason to keep the bot's script narrow and to read its transcripts.

A website visitor is a prospective client from the first message

Confidentiality duties reach beyond signed clients. The New York State Bar Association's 2026 AI guidelines state that those duties cover information from current, former and prospective clients, and that an AI tool with cross-client or cross-matter access can create conflicts or defeat ethical screens. A visitor describing a problem in a chat window is a prospective client, so the chatbot's data handling deserves the same attention as your case files.

The same guidelines recommend reasonable efforts to assess a tool's reliability, confidentiality and security, including a review of vendor terms. The safeguards they list include access controls, authentication, encryption, retention and deletion controls, limits on training with your data, subprocessor controls, incident response and audit logs.

The State Bar of Texas takes a similar line, naming client intake and chatbots among possible practice uses of AI and recommending that firms work with vendors on confidentiality, auditing, monitoring, deletion policies, staff training and AI governance.

Someone at the firm owns the script, and a person is always one click away

The NYSBA guidelines say AI may support scheduling, deadline management, drafting and other tasks, but material output must be independently verified before anyone relies on it or passes it on. For a chatbot, that means supervision: one person owns the script, reads transcripts and decides when the bot's behavior changes. A visitor should always have a visible path to a human, whether a phone number, a callback request or a staff handoff.

Judge a chatbot by how it fails, where its data goes and what it connects to

Fluency is the least useful way to judge a chatbot, since nearly every modern bot sounds smooth in a demo. Put these questions to any provider in writing before you sign.

Question to settle What a sound answer looks like
What can the bot say?A defined scope limited to firm information and intake, with legal conclusions and outcome predictions blocked
What happens when it does not know?It admits the limit and routes to a person instead of improvising
Where does the data go?Clear retention and deletion terms, access controls, and a written statement on whether your data trains the system
How does it connect to your tools?Intake details reach your CRM, and consultations book against real calendar availability
Who sees transcripts?You do, with the ability to review and adjust the script
How is it tested?A written plan for testing before launch and on a regular schedule afterward

Leads that land in an unwatched inbox cost money and sign no one

A chatbot that drops leads into an inbox nobody checks is a cost, not an asset. Decide first where intake information should land and who responds. If you have not settled on a system for tracking prospective clients, our guide to choosing a CRM for a law firm is a sensible place to start, and our overview of law firm marketing automation shows how follow-up can run without adding manual work.

One clean test before launch does not last, because the models keep changing

The ABA's guide recommends testing a firm chatbot before deployment and regularly after, deliberately trying to induce errors, and asking the provider to show its testing plan. It stresses that testing has to be ongoing because AI models evolve. A simple routine for a small firm:

  • Before launch: have two or three staff members role-play common and awkward visitors, including someone angry, someone asking whether they have a case, and someone describing an emergency.
  • Weekly at first: read a sample of transcripts and note anything the bot should have escalated.
  • Monthly after that: review booked consultations, abandoned conversations and handoff speed, and update the script.
  • After any change: retest, whether the change was yours or the provider's.

Count booked and attended consultations, not chat volume

Pick a few numbers that connect to consultations, not just conversations.

  • Conversations started versus conversations that produced contact details.
  • Consultations booked from chat, and how many were attended.
  • Response time from handoff to a human contact on urgent inquiries.
  • Escalations: how often the bot passed a visitor to staff, and whether those handoffs were the right call.

Compare those figures to your own baseline from the same period before launch. Avoid borrowing benchmarks from other firms' marketing claims; your practice area, market and ad mix will shape your numbers. If you are also watching how AI is changing the way people find firms in the first place, our piece on AI Overviews and law firm search covers the discovery side.

The chatbot is only as good as the person who picks up the handoff

A visitor does not want a clever bot. They want to know the firm is reachable, that someone competent will respond, and that the next step is simple. A chatbot can cover the hours your office is closed and take routine questions off your staff's plate, but the consultation still depends on how fast a person follows up.

So build around limits as much as capability: a defined scope, a clear AI disclosure, screening questions you control, hard stops for sensitive matters, and a clean handoff with the conversation summary attached. Then put one person in charge of reading transcripts every week until the bot earns a monthly review.

FAQ

Can an AI chatbot give legal advice to my website visitors?

It should not. A firm chatbot should share information about your firm, collect intake details and schedule consultations. ABA guidance warns against chatbots that mislead people with legal conclusions or guarantees of success, so the script should send any assessment of a visitor's situation to a person.

Do I have to tell visitors they are talking to an AI?

Disclosure is the safer practice and is recommended in the ABA's article on AI in law firm marketing. Rules vary by state, so check your state bar's current guidance, and put the disclosure in the first message of the conversation.

What should an AI chatbot collect during intake?

Only what staff need to follow up: name, contact preference, practice area and a short description of the matter, plus screening questions such as whether the person already has an attorney. Collecting less lowers your confidentiality exposure. Review the provider's retention and deletion terms before launch.

How often should I test the chatbot?

Before launch, then on a regular schedule. The ABA guide says testing should be ongoing because AI models change over time. Read transcripts weekly at first, then monthly, and retest after any change to the script or the provider's system.

Will a chatbot replace my front desk or intake staff?

No. It covers after-hours inquiries and routine questions, and your team handles the conversations that need judgment, empathy and professional responsibility.

Sources

← Back to The Forge