AI Chatbots for Law Firm Websites: What's Compliant and What Converts

16 August 2026·6 min read

Quick answer: A compliant law firm chatbot handles triage, not advice — think booking calls, answering logistics questions and collecting basic intake details, then handing off to a human before anything resembling legal advice gets said. The line gets crossed the moment a bot answers a specific legal question, comments on case merits, or lets a conversation drift into something that reads like a client relationship. Below is the traffic-light framework we use with law firm clients, three worked examples, and the guardrails that keep both your compliance and your conversion rate intact. ✨

Every law firm we talk to wants a chatbot for the same reason: after-hours enquiries turn into after-hours bookings instead of after-hours bounces. Fair enough — a good one lifts intake volume without anyone lifting a finger at 9pm on a Tuesday. But we've also seen firms bolt on an off-the-shelf AI widget, point it at their website content, and effectively let it freelance legal opinions to strangers. That's not a marketing win, it's a professional conduct risk wearing a chat bubble. 💜

What most firms get wrong

Letting the bot “help” too much. A generic AI chatbot trained on your website copy will happily answer “do I have a case?” because it's optimised to be helpful, not to know where its authority ends.
No clear handoff point. Without a defined moment where the bot stops and a human takes over, every conversation is a judgement call made by a script, not a solicitor.
Treating chat transcripts as disposable. The transcript is a record of what was said, what wasn't, and whether anything resembling advice or a retainer discussion happened.
No conflict check before information gathering. A chatbot that asks “tell us about your matter” before checking who's on the other side can walk a firm straight into a conflict.

The traffic-light framework

Green — the bot can handle this on its own: opening hours, locations, parking, fee structure in general terms (“we offer fixed-fee conveyancing”), what documents to bring, booking a consultation.

Amber — the bot collects information, it doesn't respond to it: name, contact details, a one-line description of what they need help with, the other party's name (for a preliminary conflict check), preferred callback time. No follow-up questions that probe the substance of the matter.

Red — hand off immediately, no bot commentary: anything starting with “should I”, “what are my chances”, “is this legal”, specific facts about a dispute, or a request for advice. The bot's only job here is one line: “That's a question for one of our lawyers — I've flagged this as urgent and someone will call you back shortly.” No answer, no guess, no “generally speaking”.

Worked examples

Family law firm (Southport): A prospective client messages at 10pm: “my ex won't let me see the kids, what can I do?” The bot doesn't touch it. It replies: “That sounds urgent and I want to make sure you get the right advice — can I grab your name and best contact number so one of our family lawyers can call you first thing?” It collects the basics, flags the enquiry as priority, and stops there. No mention of parenting arrangements, no reassurance about outcomes.

Conveyancing firm: Most of this firm's after-hours chat is genuinely green-light stuff — “do you handle off-the-plan purchases”, “what's your fixed fee”, “can I book a call this week”. The bot answers those directly from an approved FAQ list and books the call through the firm's calendar link. Anything about a specific contract clause gets bounced to a human with “I'll get one of our conveyancers to look at that with you.”

Immigration law practice: Visa questions are almost never simple, so this firm configured its bot to treat nearly everything as amber or red — it collects visa subclass, country of origin and a callback time, but won't attempt to answer “will I be approved” or “how long will this take” even in general terms, because those answers depend entirely on individual circumstances.

How to actually set this up

Most firms don't need a custom-built AI system — a well-configured widget on a decent chatbot platform, with a tightly written script and a locked-down knowledge base, does the job.

Restrict the knowledge base. Don't let the bot pull answers from your full website — feed it a curated, approved FAQ list so it can't improvise from an old blog post.
Build in trigger phrases. Words like “advice”, “case”, “chances”, “sue” or “liable” should force an immediate handoff regardless of context.
Only promise a callback window the firm can actually hit.
Log everything and review a sample of transcripts regularly.
Document a conflict-check step before any substantive information is collected, even informally.

💡 Heads up: a chatbot transcript can become a record like any other client communication. If your bot's default behaviour is to be maximally “helpful” rather than deliberately limited, you may be creating documentation you didn't mean to create. Configure it to under-answer, not over-answer.

Mistakes to avoid

Turning on an AI chatbot's default settings without reviewing what it will actually say to a “do I have a case” question.
Using the bot to quote specific fees or timeframes that vary matter to matter.
Letting the bot use language that reads as a guarantee of outcome.
Skipping the conflict-check step because it feels like friction — it's the step that protects the firm.
Never testing the bot yourself with the tricky questions a stressed, googling person would actually type.

Please note: general information, not legal advice — check current official guidance before relying on it.

Frequently asked questions

Can a chatbot legally give legal information, just not legal advice?

There's a real distinction between general information (opening hours, fee ranges, what a document is used for) and advice tailored to someone's specific situation, and a well-configured bot can safely stay on the information side. But that line gets blurry fast, and exactly where it sits can depend on your jurisdiction and your regulator's current guidance — worth checking with your law society rather than assuming your chatbot vendor has already thought it through.

Does an AI chatbot actually improve conversion, or is it just a nice-to-have?

For firms getting meaningful after-hours traffic, a well-scripted bot genuinely captures enquiries that would otherwise bounce, since people rarely call a law firm outside business hours but will happily type a message. The lift is usually in lead volume rather than lead quality, so your intake team still needs to filter what comes through.

Should the chatbot ever mention the firm's fees?

Only in general, published terms you'd put on a fee schedule — a fixed-fee range is fine, quoting a number based on details the bot just collected is not, because that starts to look like advice on the client's specific matter.

Who decides whether our chatbot script is compliant?

We can help with the marketing and conversion side of this, but what counts as advice, what creates a retainer, and what your professional indemnity insurer expects sit with your own practice's risk processes and your law society's current guidance, not with a marketing agency.