AI for Attorney Marketing: Content and Intake Follow-Up Within the Rules

A law firm can use AI to draft practice-area pages and timely enquiry replies if a lawyer checks every legal claim, outcome implication and advertising requirement before publication or sending. Keep intake automation to acknowledgement and routing until conflicts and confidentiality are assessed. Apply the advertising rules of every regulator governing the firm, because the details vary.

The useful boundary is between preparing words and making a professional decision. AI can turn a lawyer-approved service outline into readable copy, but it cannot decide whether a claimed specialism is permitted, whether a testimonial needs consent, or whether a prospective client's facts create a conflict. Put those decisions on an approval sheet with a named lawyer.

Follow me on Instagram@sagnikteaches

Choose one practice page and one enquiry route

Start with a single practice area that the firm actually handles and a single enquiry channel, such as the website form. Avoid replacing the whole website or connecting an AI assistant to every inbox on day one. One page gives the lawyer a manageable set of claims to check; one form lets the intake team see exactly what data arrives and who receives it. Record the baseline: page visits, submitted forms, qualified consultations and enquiries that received a human response. A high form count can hide poor-fit work or messages that never reached a lawyer.

Connect on LinkedInSagnik Bhattacharya

Write a factual source sheet before prompting. Include the services offered, the people authorised to provide them, the firm's approved biography wording, the intake hours and the appointment route. Add the correct responsible-lawyer contact details. Leave success rates, typical awards, speed promises, certifications and comparative claims blank unless the firm can support and legally use them. The ABA's Model Rule 7.1 bars false or misleading communications; it is a starting point for review, not a substitute for the rules that apply to this firm.

Subscribe on YouTube@sagnikteaches

For example, a small family-law practice might supply: divorce consultations, mediation advice, two named lawyers and an online booking link. It should not feed the model an old homepage that says the firm is the leading local practice. The new draft may repeat that unsupported superlative with smoother wording. The source sheet should be narrower than the marketing archive, because a model asked to imitate old copy may preserve old errors.

Draft a page from approved facts, then mark every claim

Give the model a bounded job: write a clear page explaining who the service is for, what happens at an initial consultation and how to enquire. Require a list of statements that need legal verification. Do not ask it to research current law from memory. If a practice page includes legal information, have a lawyer supply or verify the relevant rule, its date and its scope. A page about a procedure can become misleading when it describes an exception as universal or silently carries a rule from another jurisdiction.

Draft a practice-area page using only the approved facts below.
Audience: a person deciding whether to contact this firm.
Approved facts: [paste the lawyer-approved service sheet].
Do not invent results, credentials, deadlines, fees or client stories.
Mark each sentence that needs a lawyer's legal or advertising review.
End with a neutral enquiry invitation, not a predicted outcome.

Keep the draft and the approved source sheet together. A typical first draft (illustrative) might read, “Our team will protect your assets and resolve your divorce quickly.” The editor should replace it with a verifiable description such as, “A lawyer can discuss the documents and options relevant to your circumstances at an initial consultation.” The first sentence promises a result and a timetable; the second describes the service. The point is not to make every sentence timid. It is to make each claim defensible.

Use a line-by-line review with five marks: fact supported, legal explanation checked, advertising language checked, confidentiality checked and next action clear. A line can pass four and still fail the fifth. “Board-certified specialist” is a factual and regulatory claim, while “we respond within one working day” is an operational claim that needs a measured service standard. Model Rule 7.2 lets a lawyer claim certification as a specialist only when an approved or accredited body did the certifying and is named in the communication, and it requires advertising to carry the name and contact details of at least one lawyer or firm responsible for its content. Verify the local version before using either claim.

Keep page history. Record the prompt version, source-sheet date, lawyer approver and publication date. When a rule or practice changes, the firm can find affected pages. A shared table with page URL, owner, last legal review and next review date is sufficient for a small firm. The claim-checking method for AI marketing copy helps with general marketing evidence; law firm pages also need a practising lawyer's regulatory review.

Handle client stories as consent projects

Client stories, reviews and case examples create a separate risk. A name removed from a story may leave a recognisable combination of dates, dispute type, role or unusual facts. A public review does not automatically authorise the firm to reveal its side of a matter; the ABA's FAQ on advertising in the digital age notes that a client's negative public review generally does not waive the lawyer's duty of confidentiality. Model Rule 1.6 protects information relating to representation, and Rule 1.18 covers information from prospective clients even where no engagement follows. Obtain the specific permission required by the applicable rules, then ask a lawyer to decide what may appear.

A workable consent record says exactly what text and image will be used, where it will appear, how long the permission applies and whom the client can contact about a change. The firm should retain the version the client approved. If the client approved a short website quotation, do not ask AI to turn it into a detailed social post. Every added fact needs a new review. Some regulators impose further testimonial or results-related conditions, so the marketing team must check those before scheduling publication.

Consider an illustrative client quotation: “They kept me informed during a difficult dispute.” A draft expansion saying “We won the case in six weeks and recovered every cost” invents both a result and a timeframe. The safe edit is to preserve the approved quotation exactly, subject to consent and regulator review. If consent is absent, use a generic process example clearly labelled illustrative: “A client with an urgent deadline would first speak to an intake coordinator, then a lawyer if the matter is suitable.”

Make consent a gate in the publishing workflow: no record, no story. That simple rule is easier to apply than debating whether an account is sufficiently anonymised after the model has already produced a compelling draft. The claim evidence checklist is useful for non-client claims, but it cannot replace consent or a confidentiality decision.

Design an intake form that asks for less

Ask for the minimum information needed to route an enquiry: contact details, broad matter type, a brief timing question and preferred contact method. If the firm needs party names for conflict screening, explain why and send them through the approved route. Avoid an open prompt asking visitors to upload every document or explain every sensitive detail before the firm has decided whether it can speak with them. Intake data may be confidential, and it often contains facts about other people.

Show a clear notice at the form, checked by the firm's lawyer, about what submitting it does and does not establish. The notice must fit the actual process: a sentence that no lawyer-client relationship is formed by submission cannot be used to excuse careless handling of prospective-client information. The ABA's Model Rule 1.18 is a useful reference for the duties owed to a prospective client. Local rules and the firm's conflict policy decide the exact intake design.

If a chatbot fronts the form, treat it as part of the firm's advertising and intake, not a neutral widget. The same ABA FAQ points out that a lawyer's supervisory duty under Model Rule 5.3 extends to AI tools such as a chatbot, so the lawyer is responsible for what it says. It recommends reviewing every script the bot can use and showing a prominent notice that it is automated and cannot give legal advice. In practice, export the bot's full answer library before launch and have the responsible lawyer mark each answer approved, rewritten or removed. A bot that answers “Do I have a case?” with anything other than a route to a person has failed that review.

Route form fields into a restricted intake record before an AI tool sees them. Use an approved business tool with terms and settings reviewed for client information; give staff role-based access. For drafting a generic acknowledgement, the model needs the matter category and next action, not the person's narrative or attachments. Test the form with made-up entries first. A personal-injury enquiry submitted under “employment” is a good test of whether a routing assistant flags uncertainty instead of silently sending it to the wrong team.

Write follow-up emails that do not give advice

The first email should confirm receipt, state a realistic next step and identify urgent limits. It should not diagnose the case, suggest a legal strategy or promise that the firm will take it. Use variables for first name and broad matter category, then a coordinator or lawyer checks the version before activation. If the team cannot reliably answer within the stated period, do not put that period in the template. A delayed but truthful expectation is better than a polished promise the firm misses.

An illustrative acknowledgement reads: “Hello [first name], thank you for contacting the firm about your enquiry. Our intake team will review the information you submitted and tell you whether we can arrange a conversation. We have not agreed to represent you. Please do not send further sensitive documents until we request them through our secure process. If your matter has an immediate deadline, seek prompt legal help rather than waiting for this email.” A lawyer must adapt that wording to the firm's rules and service.

For a second message, send a practical request only after the intake team has checked the route: “Hello [first name], to arrange the next step, please confirm a suitable time and the names of the parties for our conflict check using the secure form linked below.” A poor AI draft might say, “Your matter looks strong; upload your full file and we will begin.” Delete the assessment and the implied acceptance. The legal intake call guide covers similar boundaries when a voice assistant receives the first contact.

Set a boundary on follow-up frequency. If a prospective client submits a form and does not reply to one request for a suitable time, a single reviewed reminder may be appropriate; repeated promotional emails can create a different advertising and consent question. Keep the enquiry follow-up separate from a newsletter subscription. The same form should not quietly add people seeking legal help to a marketing list. Record any subscription choice distinctly, including its date and the wording shown at the time.

Now suppose the contact system proposes “You have three days left to claim.” That line may be plausible for one kind of matter but dangerous across a mixed intake queue. Even a lawyer-approved template cannot infer a deadline from an unchecked form category. Replace it with “If you are concerned about a deadline, seek legal advice promptly.” Where the firm has a verified deadline in a specific matter, a lawyer should handle the response. Keep an exception queue for messages mentioning court dates, urgent safety concerns or pending expiry, and test that they reach a person.

Do not use one automated sequence for every enquiry. A declined matter needs a reviewed non-engagement message; a conflict hit needs a restricted human route; an existing client needs its own record and contact rules. An inbox rule can mark these categories for staff without generating a legal response. Review every branch with test messages, including a duplicate enquiry, an unreachable address and a message that mentions an imminent hearing. Automatic sending should be limited to text the lawyer has approved for that exact branch.

Run a 30-enquiry pilot with named approvals

Illustrative numbers make the decision clearer. Suppose a four-lawyer firm receives 30 website enquiries in a month. Twelve are clearly outside its practice area, five need conflict review, eight are potentially suitable and five lack enough information. Before the pilot, staff spend about eight minutes finding and sending each basic acknowledgement: 30 times eight is 240 minutes. A controlled template might reduce the mechanical portion to three minutes per enquiry, or 90 minutes, but only if the team still reads and routes each message. The 150-minute difference is an illustration to test, not a promised saving.

Build three approved acknowledgement variants and a routing table. For the twelve out-of-scope enquiries, staff choose the reviewed non-engagement route. For the five conflict candidates, no AI answer is sent beyond a neutral receipt; a lawyer decides the next message. For the eight potentially suitable matters, intake offers a conversation without saying the case is strong. For the five incomplete forms, the team asks only for the missing routing information. Record the handling time and any correction to an AI suggestion. A branch that saves typing but doubles corrections is a poor branch.

At the same time, revise one practice-area page. Count its claims: suppose there are 18 substantive statements, with four legal explanations, two credentials, one service-time claim and one testimonial. The lawyer checks each against a source or record. If three cannot be supported, remove or rewrite them before publication. The resulting page may be shorter than the AI draft, but the approval record is stronger. Keep the old page available internally so the firm can compare enquiries and avoid attributing every change to AI.

Measure qualified consultation requests, correct routing, response time and correction rate over the pilot. “Qualified” should mean a category the firm might handle after conflict and capacity checks, not simply any submitted form. Review at least a sample of the actual outgoing messages weekly. If the automation incorrectly classifies sensitive enquiries, stop that branch and return it to human triage. A firm can keep AI-assisted drafting even if automatic sending proves unsuitable.

Keep a simple error log: date, branch, suggested text, human correction and reason. An illustrative entry might read, “15 October; incomplete enquiry; model asked for medical records; coordinator removed request; only appointment availability was needed.” If several entries show the same failure, change the approved prompt or remove that branch. Do not treat the number of messages sent as proof of improvement. A faster wrong reply can damage trust and produce extra staff work. Compare median response time alongside correction rate and the number of suitable consultations actually booked.

Check the page and follow-up together. The practice page may promise a call back while the template offers only a booking link; that mismatch invites complaints even if both sentences are individually lawful. Read the public page as a new visitor, submit a test form and follow every resulting email. Then repeat with a test enquiry outside the firm's practice area. This end-to-end test catches unreviewed copy in form confirmations, calendar notices and automated reminders, where marketing teams often forget that the same advertising and confidentiality concerns still apply.

Build a publication gate the team can actually use

Assign a page owner, intake owner and lawyer approver. The page owner checks readability and links. The intake owner checks that routing, timing and templates match operations. The lawyer checks legal explanation, advertising wording, consent and confidentiality. For each release, save the final text, approval date and the regulator rule checked. The applicable bar rules may differ across jurisdictions and may change; verify them with the regulator rather than relying on a model or a generic checklist.

The release sheet can be short: “Page: family-law consultation; approved facts updated: 12 October; claims checked: 18 of 18; client story consent: none used; responsible lawyer: [name]; intake branch tested: three sample messages; next review: [date].” This is an illustrative completed record, not a legal form. A blank “approved” tick without the page version or approver is hard to audit when a sentence changes later.

Review the page again when staff, services, fees, contact details or relevant rules change. Check live forms after publication, not only the staging copy. A broken booking link can turn a compliant page into a poor intake experience; a copied testimonial can reintroduce a consent problem. The marketing automation setup guide can help with a wider rollout after this narrowly controlled route proves reliable.

Questions before a law firm publishes AI-assisted copy

Can an AI tool answer a prospective client's legal question automatically?

Use automation for acknowledgement, scheduling and requests for limited information. A lawyer should decide whether to provide a legal answer after conflict screening and reviewing the facts. A scripted reply should explain that no representation has been agreed and should not predict an outcome. Check the applicable regulator's rules and the firm's intake procedure before enabling any automatic response.

Does a client review give the firm permission to publish case details?

No. A public review is not a blanket waiver of confidentiality. Obtain informed permission for the specific details and channel, remove unnecessary identifiers, and have the supervising lawyer review the result. If permission is uncertain, publish a general description of the service instead of a client story.

Must every AI-written practice page carry an AI label?

The central review is whether the page is accurate, appropriately qualified and compliant with the rules where the firm practises. Requirements for labels and disclaimers vary. Ask the lawyer responsible for advertising to check the relevant regulator's current rule and the final page. An AI label does not cure a misleading claim.

Further reads

• AI Marketing for Small Law Firms: Content, Reviews and the Rules — Review the wider law firm marketing controls.
• How to Write an AI Content Policy for Your Marketing — Turn review responsibilities into a written content policy.
• How to Turn One Customer Story Into a Month of Marketing With AI — Plan reuse only after the client approves the story.
• Best AI Chatbot for Lawyers: General Models vs CoCounsel — Compare intake chat boundaries with legal-answer tools.
• How to Measure Whether AI Is Improving Your Marketing Results — Measure qualified enquiries rather than page volume.
• AI Content for Law Firms: Website Articles That Pass Advertising Rules — A topic brief, drafting prompt, twelve advertising-rule checks and a before-and-after rewrite for law firm website articles.
• AI for Attorney Research: A Citation-Checking Workflow — A repeatable way to turn AI research suggestions into checked authorities before a draft reaches the court.
• AI for Franchise Lead Generation: Staying Inside Advertising Rules — AI for franchise lead generation without breaking advertising or earnings-claim rules: prospect ads, landing pages, FDD timing and lead handling.
• Can You Use AI to Advertise a Professional Practice Within the Rules? — How a regulated practice can draft adverts with AI and still meet professional advertising rules, with a compliance check you can copy.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ABA Model Rules 7.1, 7.2, 1.6, 1.18 and 5.3; ABA FAQs on Attorney Advertising and Ethics Rules in the Digital Age. Check the applicable bar regulator's current rules before publication.