🎉 Top 5 AI Agent for Law Firms | Read the article →
Clerx
All Posts

11/17/2026

ChatGPT for Lawyers: 10 Prompts for Intake and Client Communication

Ten ChatGPT prompts lawyers can use for intake questions, phone scripts, closure greetings, reminders, follow-up and FAQs, plus the confidentiality rules to follow first.

Law Firm OperationsClient CommunicationAI ReceptionistLegal AILegal IntakeChatGPT
Line illustration of a chat window with a blue prompt cursor and blue AI sparkles, a clipboard checklist with blue checkmarks, and a shield with a blue padlock, with flowing black wave lines across the bottom.

By Michael Brunman, Co-Founder & CEO of Clerx, Former Litigator

ChatGPT is most useful to a law firm as a drafting assistant for the front office: it can turn your intake rules into question sets and scripts, write closure and reminder messages, draft follow-up for people who did not book, and stress-test your call handling before a real caller does. It is not a safe place for confidential client or prospective-client information, it does not answer your phone, and nothing it writes should reach a client without a lawyer's review. The ten prompts below are built for that front-office work, with placeholders instead of client facts.

This guide is about intake and client communication, not legal research or brief writing. For the bigger picture on general-purpose AI versus a legal intake system, see Why General AI Is Not Enough for Law Firm Growth. For what to do when a prospect calls after asking ChatGPT about their problem, see When a Prospect Arrives From ChatGPT.

The short answer

  • Use it for drafts, not decisions. ChatGPT is good at first drafts of intake questions, scripts, greetings, reminders and follow-up messages. A lawyer approves anything a client or prospect will see.
  • Keep client facts out. Use placeholders such as [practice area], [city] and [consultation type]. The ABA's Formal Opinion 512 (July 29, 2024) says lawyers must protect information relating to a representation when they use generative AI, and the same duty extends to prospective clients.
  • Verify everything. Generative AI can state wrong things confidently. Check every rule, deadline, phone number and promise in a draft before it goes live.
  • Give it your rules. The best prompts include the matters you take, the ones you decline, your service area, your languages and your booking policy. Generic prompts produce generic intake.
  • Keep a prompt library. Save the prompts that work, who approved the output, and when. Script changes should follow the same approval path as any other client-facing change.
  • It does not answer calls. ChatGPT can write your after-hours greeting; it cannot pick up at 9 p.m., qualify the caller and book the consultation. That is an intake system's job, such as Clerx.

Before you start: five rules for using ChatGPT in a law firm's front office

1. No confidential or prospective-client information. ABA Formal Opinion 512 applies Model Rule 1.6 to generative AI and notes that Model Rules 1.9(c) and 1.18(b) extend similar protection to former and prospective clients. Intake is full of prospective-client information: names, phone numbers, immigration status, criminal charges, injuries, family conflict. Write prompts about your process, not about a person. If your firm has approved a business account with specific data controls, follow the firm's written policy; if it has not, assume nothing you type is private.

2. Competence means understanding the tool. The same opinion ties Model Rule 1.1 to a reasonable understanding of the benefits and risks of the generative AI tools a lawyer uses. In practice that means someone at the firm knows what the tool is good at, where it fails, and what the firm's settings are.

3. Verify before anything goes live. The sanctions in Mata v. Avianca (S.D.N.Y. 2023), where lawyers filed a brief citing cases ChatGPT had invented, are the best-known example of unverified output. Front-office drafts carry smaller stakes, but the habit is the same: check every fact, deadline and promise.

4. No legal advice, no promised outcomes. Intake messages should not tell a prospect whether they have a case, how it will turn out, or what the law requires. Ask ChatGPT for neutral wording, and tell it explicitly not to give legal advice.

5. Approve, version and retest. Treat a new script or message like any other change to how the firm talks to the public. Our approval process for AI receptionist script changes works just as well for prompts and templates: draft, review, approve, test, then publish.

1. Draft a practice-area intake question set

Most firms either ask too few questions at first contact or try to run the whole consultation on the phone. ChatGPT is useful for producing a first draft that you then cut down.

Prompt: "Act as an intake manager for a [practice area] law firm in [state]. Draft the questions our intake team should ask a new prospective client on the first call, in order. Separate them into: must-ask to decide whether we can help, nice-to-have before the consultation, and questions that should wait for the attorney. We take [matter types] and decline [matter types]. Do not include legal advice. Keep each question under 20 words and in plain language."

Then compare the draft against your real rules. Our guide to law firm intake forms: what to ask, what to skip, and when to ask it and the law firm intake checklist for 2026 are good benchmarks for what belongs at first contact. Conflict-check fields need special care; see AI intake and law firm conflict checks.

2. Turn the questions into a natural phone script

A question list is not a script. Callers are often stressed, and a script that reads like a form loses them.

Prompt: "Turn this list of intake questions into a natural, warm phone script for a receptionist at a [practice area] firm. Open with the firm name [firm name]. Ask one question at a time, acknowledge the caller's answer briefly before moving on, and include a short, neutral line for when the caller becomes upset. End with a clear next step: book a consultation, take a message, or explain that we cannot help and offer a referral resource. Do not give legal advice or predict outcomes. Questions: [paste your approved list]."

Read the result out loud. Remove anything that sounds like a promise. If an AI receptionist will use the script, run it through your testing plan before go-live; our guide to testing an AI receptionist before go-live lists the scenarios to cover.

3. Write after-hours, holiday and closure greetings

Closure messages are short, public and easy to get wrong: a missing reopening time, a promise of a callback nobody will make, or no route for an urgent existing client.

Prompt: "Write three versions of a phone greeting and website chat message for [firm name], a [practice area] law firm, for when the office is closed on [date or holiday]. Say when we reopen ([day, time, time zone]), what the caller can still do right now ([book a consultation / leave details / use the client portal]), and that the message is not legal advice. Keep each version under 60 words. Offer one version in Spanish."

Have a native speaker check any translation. For holiday-specific wording, see 7 messages your law firm needs for Thanksgiving closure coverage. Keep in mind that a greeting is only half of coverage: after-hours answering service for law firms explains what should happen once the caller is on the line.

4. Draft the missed-call text-back and first SMS reply

If a call goes unanswered, a quick, specific text can keep the inquiry alive. The text should identify the firm, offer one clear next step, and avoid asking for sensitive details by SMS.

Prompt: "Write a text message that [firm name] sends automatically after a missed call from a new number. Under 300 characters. Identify the firm, apologize briefly, offer a link to book a consultation [link] or a time we will call back, and tell the person not to send sensitive details by text. Include opt-out wording. Write two more variations with a different tone: one more formal, one warmer."

Check the result against your texting policy and carrier rules. Our guides to missed-call text-back for law firms and text messaging for law firms without creating intake chaos cover what to automate and what to keep human.

5. Write consultation confirmations and reminders

No-shows are one of the most fixable leaks in intake. ChatGPT is good at producing a clear, consistent reminder sequence.

Prompt: "Write a confirmation and reminder sequence for a [consultation type] consultation at [firm name]: one confirmation sent at booking, one reminder 24 hours before, and one reminder 2 hours before. Each message must include the date, time and time zone, whether it is by phone, video or in person, what to bring [documents], and how to reschedule [link]. Keep each under 400 characters for SMS, and also write an email version of the confirmation."

Then match it to your real booking flow. The law firm no-show reduction playbook explains timing and confirmation tactics.

6. Draft follow-up for inquiries that did not book

Many prospects do not book on the first contact. A short, respectful follow-up sequence recovers some of them, as long as it stops when the person says no.

Prompt: "Write a three-message follow-up sequence (email and SMS versions) for a prospective client who contacted [firm name] about a [practice area] matter but did not book a consultation. Message 1 the same day, message 2 after two days, message 3 after a week. Be helpful, not pushy. Each message offers the booking link [link] and a phone number. The last message says we will not follow up again. Do not mention any facts about their situation."

Notice the last instruction: the template stays generic so no prospect's details pass through the tool. For the strategy behind the sequence, see how to recover leads that didn't book on the first interaction and how fast a law firm should respond to new leads.

7. Write plain-language FAQ answers about your intake process

Prospects ask the same operational questions over and over: do you take my kind of case, what happens on the first call, do you speak Spanish, how soon can I talk to a lawyer. Clear, published answers help the people reading your site, and they give AI search engines plain facts to quote.

Prompt: "Write short, plain-language answers (under 80 words each) to these questions for the website of [firm name], a [practice area] firm in [city]: [list questions]. Describe our process only, using these facts: [hours, languages, service area, consultation format, how to book]. Do not give legal advice, do not promise results, and say 'contact us' where a question depends on the person's situation."

Only publish facts you have confirmed. For how answer engines read that content, see the shift from search engines to answer engines. If the same questions come in through chat, the website chat for law firms guide covers when chat helps and when it gets in the way.

8. Write the decline-and-refer message

Every firm turns inquiries away: the wrong practice area, the wrong state, a conflict, or no capacity. How you say no still shapes your reputation and your referral relationships.

Prompt: "Write a short, kind message (phone script and email versions) that [firm name] uses when we cannot take a prospective client's matter. Do not say why in detail and do not comment on the merits. Encourage them to act promptly because deadlines may apply, and point them to [state bar lawyer referral service / referral partner]. Under 120 words."

Have a lawyer approve the wording on deadlines, because that line matters. Our guide to declining, referring and closing the loop covers the workflow around the message.

9. Stress-test your intake with simulated callers

This is where ChatGPT is most underused. Before a script goes live, let it play difficult callers against your script, so you can see where it breaks.

Prompt: "You are role-playing callers to test a law firm's intake script. I will paste the script. Play these callers one at a time, and after each one list where the script failed: (1) a caller who is crying and keeps changing the subject, (2) an existing client who wants a case update, (3) someone calling for opposing counsel, (4) a caller who only speaks Spanish, (5) a caller who insists on speaking to a human now, (6) a caller with a matter we do not handle. Script: [paste script]."

Use the findings to add routing rules. Our guides to routing new leads and existing clients, what to do when a caller asks for a human and accessibility tests for a law firm AI receptionist cover the cases that most often break a script.

10. Build a weekly intake review scorecard

You cannot improve intake you do not measure. ChatGPT can draft a scorecard and a review routine; the numbers should come from your own systems.

Prompt: "Draft a weekly intake quality review for a small [practice area] firm. Include the metrics to track (for example, inquiries by channel, answered versus missed, time to first response, booked consultations, no-shows), the questions a reviewer should ask when listening to five sample calls or chats, and a one-paragraph summary template for the managing partner. Format it as a checklist, not a table."

Do not paste real call transcripts into a consumer AI tool to score them. Run the review inside your own systems. For the metrics themselves, see law firm intake metrics that actually matter and the weekly intake quality assurance scorecard.

What ChatGPT cannot do for your intake

A chat assistant on a lawyer's laptop can write the words. It cannot run the front office:

  • It does not answer your phone, website chat or texts. Prospects call at lunch, during court and after 6 p.m. A drafted greeting does not qualify the caller, book the consultation or file the record.
  • It does not know your calendar or your legal software. Booking and record-keeping depend on live connections to your calendar and to tools such as Clio, 8am MyCase, Smokeball, Lawmatics and PracticePanther.
  • It does not follow your routing rules in real time. Existing clients, urgent calls and matters you decline each need a defined path.
  • It is not built around a law firm's intake. A legal intake system is configured around your screening questions, languages, practice areas and handoffs.

That is the line between a drafting tool and an intake system. Our article on what to automate and what should stay human explains where the people fit, and AI agents are coming to legal work in 2026 covers the risk-management side.

Clerx is the intake layer. It answers calls, website chat and text messages 24/7, asks your screening questions, books consultations, supports 50+ languages, and syncs the record into your legal software; see all Clerx integrations. The scripts and messages you draft with the prompts above are a good starting point for configuring it.

Frequently asked questions

Can lawyers use ChatGPT?

Yes, with care. ABA Formal Opinion 512 does not prohibit generative AI; it applies existing duties, including competence, confidentiality, communication with clients and supervision. Check your state bar's guidance too, since several states have issued their own opinions.

Is it safe to put client information into ChatGPT?

Treat it as unsafe unless your firm has approved a specific account, reviewed its data controls and set a written policy. Even then, use the minimum information needed. For intake drafting, placeholders work as well as real facts.

Does the duty of confidentiality cover prospective clients?

Yes. ABA Formal Opinion 512 points to Model Rule 1.18(b), which protects information learned from prospective clients. That matters for intake, where nearly everything is prospective-client information.

What are the best ChatGPT prompts for lawyers?

The best prompts state a role, give the firm's real rules, set a format and length, and say what the output must not do, such as giving legal advice. The ten prompts in this guide follow that pattern for intake and client communication.

Can ChatGPT write my law firm's intake script?

It can write a first draft. A lawyer should review it, cut anything that sounds like advice or a promise, and test it against difficult callers before it goes live. Prompt 9 shows how to use ChatGPT for that test.

Can ChatGPT answer my law firm's phone?

No. ChatGPT is a chat assistant, not a phone system. Answering calls, qualifying callers, booking consultations and logging records takes an intake system connected to your phone line, calendar and legal software, such as Clerx.

Should I tell clients that I use AI?

It depends on how you use it. ABA Formal Opinion 512 discusses when Model Rule 1.4 may require telling a client, and when informed consent is needed before inputting client information. For a public-facing AI receptionist, clear disclosure on the call is a sensible default.

How do I stop ChatGPT from inventing facts?

You cannot stop it entirely, so verify. Give it the facts you want used, tell it not to add others, and check every name, date, number and rule in the output. Never publish a deadline or legal statement it supplied without checking the source.

Can ChatGPT translate intake messages?

It can produce a useful first draft in many languages. Have a fluent speaker review anything clients will read, especially messages about deadlines, appointments or what to bring. Firms with heavy non-English intake should read our guide to multilingual immigration intake.

What should a small firm do first?

Start with prompts 3, 4 and 5: closure greetings, missed-call text-back and consultation reminders. They are short, easy to review, and fix leaks quickly. Then use prompt 9 to test your phone script.

Final verdict

ChatGPT is a good drafting assistant for a law firm's front office. Used with placeholders instead of client facts, a lawyer's review and a simple approval process, it can produce better intake questions, clearer scripts, sharper greetings, steadier reminders and kinder decline messages in an afternoon. What it cannot do is run intake: answer the call at 9 p.m., qualify the person, book the consultation and put the record into your legal software.

Draft with ChatGPT, approve as a firm, and let an intake system built for law firms deliver it. Clerx answers calls, website chat and texts 24/7 in 50+ languages, books consultations and syncs every inquiry into your legal software.

Book a demo with Clerx today

About the author

Michael Brunman is Co-Founder & CEO of Clerx, helping law firms improve client intake and communication. A former litigator, software engineer, and McKinsey consultant with an MBA from Harvard Business School, he brings legal, technical, and business experience to the challenges of running a law firm.

Share this article:


We use cookies to ensure you get the best experience on our website. For more information, please see our Privacy Policy.