In this guide
- Law firm chatbot ethics: what bar regulators have said
- The accidental attorney-client relationship
- Unauthorized practice and "no legal advice"
- Saying it's a bot
- Confidential information in chat
- A safe intake-only setup (example instructions)
- Supervision checklist (Model Rule 5.3)
- Frequently asked questions
A law firm can ethically use a website chatbot, but the bar opinions we've read treat it as a supervised intake tool, not a lawyer. Law firm chatbot ethics comes down to five things: say it's a bot, don't let it give legal advice, don't let it sound like it has taken the case, keep confidential details out of the chat, and review what it says. Rules vary by state, so check your own jurisdiction's rules and opinions before you switch one on.
This post is for solo and small-firm attorneys deciding whether to put a chatbot on their site. It summarizes what regulators have said, explains each risk in plain terms, and ends with example settings and a supervision checklist.
This is general information, not legal advice or an ethics opinion. We're a software company. Check your state's rules and ethics opinions, and talk to your own ethics counsel or your bar's ethics hotline before relying on any of it.
Law firm chatbot ethics: what bar regulators have said
Three sources are worth reading in full. We've linked the primary texts.
Oregon State Bar, Formal Opinion 2026-208, "Chatbots and AI Agent Communications" (approved February 2026). This is the most directly relevant opinion we've found. It asks whether Oregon lawyers may use chatbots to communicate with clients or potential clients on their websites, and answers: "Yes, qualified." It identifies the Oregon rules on competence (1.1), prospective clients (1.18), supervision of nonlawyers (5.3), false or misleading communications (7.1) and solicitation (7.3(a)), and confidentiality (1.6). It treats the chatbot vendor as a third party the lawyer must supervise, and it says lawyers "are responsible for and have ethical duties with respect to chatbots they use with clients or prospective clients."
ABA Formal Opinion 512, "Generative Artificial Intelligence Tools" (July 29, 2024). This is the ABA's general opinion on lawyers using generative AI. It isn't about website chatbots specifically, but it frames the duties that apply: competence, confidentiality, client communication, supervision, candor and reasonable fees. It notes that issues may also arise under Model Rule 7.1, and cites the Florida Bar's discussion of AI chatbots in advertising.
The Florida Bar, Ethics Opinion 24-1 (January 2024). Both the Oregon and ABA opinions cite it. As the Oregon opinion summarizes it, Florida warned that chatbots in marketing and intake carry the risk "that a prospective client relationship or even a lawyer-client relationship has been created without the lawyer's knowledge," and was one of the states to opine that a chatbot that doesn't identify itself as a chatbot could be misleading.
Also relevant is the FTC's action against DoNotPay, which marketed itself as "the world's first robot lawyer". The FTC's final order, announced February 11, 2025, required $193,000 in monetary relief and prohibits the company from advertising that its service performs like a real lawyer unless it has sufficient evidence. That case was about a consumer product claiming to replace lawyers, not a firm's website assistant. The lesson for firms is narrower: don't describe your chatbot as doing legal work it can't do.
Your state may have its own opinion, a different version of these rules, or no guidance at all yet. That's the first thing to check.
The accidental attorney-client relationship
The biggest risk isn't the bot saying something wrong. It's the bot saying something that makes a visitor reasonably believe a lawyer has taken their case.
The Oregon opinion puts it plainly: in Oregon, whether an attorney-client relationship exists depends on the client's reasonable belief, and "it is certainly conceivable that a chatbot that sounds like a lawyer could cause a prospective client to believe that an attorney-client relationship had formed." If that happens, the opinion notes, the lawyer could owe that person duties of competence, diligence and communication.
Even short of a full relationship, a person who consults with a lawyer about possibly forming one is a prospective client under ABA Model Rule 1.18. Information learned from a prospective client is protected even if no representation follows, and receiving information that could be "significantly harmful" to that person can disqualify you from acting against them in the same or a substantially related matter. The Oregon opinion says that when a chatbot gathers enough information about a prospective client, it's important to enter them into your conflicts system promptly.
What makes a bot sound like it has taken the case:
- "Yes, you have a strong claim."
- "We can help you with that. Tell me what happened."
- "Don't worry, we'll handle it."
- "Based on what you've told me, you should..."
- Sending an engagement letter or scheduling confirmation without a person involved.
What keeps it clearly on the right side:
- A plain statement, in the greeting and before the contact form, that chatting does not create an attorney-client relationship.
- Answers only about the firm: practice areas, locations, fees you publish, how to book a consultation.
- A handoff to a person for anything about the visitor's own situation.
Oregon also says it would be prudent for a responsible lawyer to be notified of chatbot interactions and to promptly review them to decide whether a relationship has formed, whether other duties have been triggered, and whether to take the matter or tell the person no relationship will follow.
Unauthorized practice and "no legal advice"
ABA Model Rule 5.5(a) says a lawyer shall not practice law in a jurisdiction in violation of its rules, "or assist another in doing so." The Oregon opinion warns that lawyers "must be careful that the chatbot does not attempt to provide legal advice, which could implicate the attorney in the unauthorized practice of law," citing a joint Pennsylvania and Philadelphia bar opinion to the same effect.
There's also a jurisdiction problem. A bot on your website answers anyone, anywhere. A question like "can I sue my landlord over my deposit?" has a different answer in every state, and your bot doesn't know where the visitor is.
The safest line is simple: the bot talks about the firm, never about the visitor's legal situation.
| Let it answer | Have it refuse and hand off |
|---|---|
| Which practice areas you handle | Whether they have a case |
| Which counties or states you practice in | What the law says about their facts |
| Consultation fees and payment options you publish | Deadlines, statutes of limitations, filing steps |
| Office hours, locations, parking | Likely outcomes or settlement values |
| Languages spoken | Questions about an existing matter |
| How to book a consultation | Conflict checks and case evaluations |
Our assistant answers only from your published pages and notes, and doesn't fill gaps from general knowledge about the law. Every conversation appears in your dashboard for review. If your pages don't cover something, it says so. Our law firm chatbot page shows the questions people typically ask and how it handles each. Because it only knows what you've published, your site's own content needs to stay accurate too. Rule 7.1 applies to the whole website, chatbot included, and our guide to stopping a chatbot making things up has a pre-launch test you can adapt.
Saying it's a bot
The Oregon opinion notes that "more than one state has opined that the use of a chatbot that does not identify itself as a chatbot rather than a person could be a misleading communication." Even where your state hasn't said so, disclosure costs nothing.
Practical steps:
- Name it as an assistant, not a person. "Smith Law assistant", not "Sarah".
- Say so in the greeting. "Hi, I'm an automated assistant for Smith Law. I can answer questions about the firm, but I'm not a lawyer and can't give legal advice."
- Don't let it overclaim. No "our AI evaluates your case", no "instant legal answers". That's the DoNotPay problem in miniature.
- Make reaching a human obvious. Your phone number and contact page should be in the bot's sources so it can give them.
Confidential information in chat
Under Model Rule 1.18, what a prospective client tells you is protected. A chat box invites people to type their whole story, including facts that could create a conflict.
Three ways to limit that:
- Tell visitors not to share confidential details. Put it in the greeting and in the text above the contact form.
- Ask for the minimum. Our lead form asks for an optional name and an email or phone number, and keeps the question the visitor typed. The attorney sees "landlord kept deposit" rather than a bare phone number, without inviting a full narrative.
- Understand where the data goes. Oregon says lawyers should understand how a chatbot stores information, whether prompts are used to train the product, and which underlying vendors the provider relies on. Read your chatbot provider's privacy policy, ours included, and ask questions before you go live.
A safe intake-only setup (example instructions)
This is what we'd suggest for a small firm using our chat assistant. Adjust the disclaimer wording with your own ethics counsel; it's an example, not approved language.
Greeting:
Hi, I'm Smith Law's automated assistant. I can answer questions about
our firm, practice areas and consultations. I'm not a lawyer, I can't
give legal advice, and chatting here doesn't make you a client of the
firm. Please don't share confidential details in this chat.
Contact form prompt:
Leave your name and an email or phone number and someone from the firm
will contact you. Please don't include confidential details. Submitting
this doesn't create an attorney-client relationship.
Standing instructions:
You answer questions about Smith Law only: practice areas, counties we
serve, consultation fees, payment options, office hours, languages
spoken and how to book a consultation.
Never give legal advice. Never say whether someone has a case, what the
law says about their situation, what they should do, what deadlines
apply, or what outcome to expect. If asked, say an attorney would need
to review their situation and offer to take their contact details.
Never say or imply that the firm has agreed to represent anyone, or that
a consultation has been booked or confirmed.
Do not ask for details of what happened. If someone starts describing
their situation, remind them not to share confidential details in chat
and offer to take their contact details instead.
Do not discuss existing cases. Tell current clients to contact their
attorney directly at the number on our contact page.
If someone describes an emergency or says they are in danger, tell them
to call 911.
Always say you are an automated assistant if asked whether you are a
person or a lawyer.
Then turn on lead capture so requests come to your inbox with the question attached. Before launch, test it yourself: ask it "do I have a case?", "what's the deadline to file?", "are you a real lawyer?", and type a long description of a fake dispute. Every answer should be a polite refusal and an offer to take contact details.
Supervision checklist (Model Rule 5.3)
ABA Model Rule 5.3 requires lawyers with managerial or direct supervisory authority to make reasonable efforts to ensure a nonlawyer's conduct is compatible with the lawyer's professional obligations. Both ABA Opinion 512 and the Oregon opinion apply that duty to outside technology providers. A checklist for a small firm:
Before launch
- Read your state's rules and any ethics opinions on chatbots, AI or lawyer websites.
- Read the vendor's privacy policy and terms. Know where chats are stored and who can see them.
- Confirm the bot answers only from material you control, and that you can see every conversation.
- Set the greeting, contact form text and instructions above, reviewed by the responsible attorney.
- Check every page the bot reads for accuracy. Outdated fees or practice areas become outdated chatbot answers.
- Run the refusal tests and keep a note of the results.
Every week
- A named attorney (or trained staff, as Oregon allows) reads the week's conversations and leads.
- Run each new lead through your conflicts process before substantive contact.
- Check the bot's unanswered questions. Our answer gaps view lists them. Add firm information where appropriate. Never add content that answers legal questions.
- Correct anything misleading at the source page, then retrain.
When anything changes
- Fees, practice areas, office locations or attorneys: update the page first, then the bot.
- A new ethics opinion in your state: re-read it against your setup.
Our pricing page lists plans; lead capture is included from Starter up, and the 3-day free trial lets you test the refusals before anything goes on your live site.
Frequently asked questions
Can a law firm use a chatbot on its website?
Generally yes, with safeguards, though rules vary by state. Oregon's Formal Opinion 2026-208 answers "Yes, qualified" and expects lawyers to understand the tool, protect confidentiality, avoid misleading statements and supervise it. Check your own state's rules and opinions; this is general information, not legal advice.
Can a chatbot create an attorney-client relationship?
Bar regulators have warned that it might. Oregon's opinion says a chatbot that sounds like a lawyer could lead a visitor to reasonably believe a relationship exists, and anyone who consults about possible representation can be a prospective client with protected information under Rule 1.18. Clear disclaimers, no case-specific answers, and prompt lawyer review reduce the risk.
What disclaimer should a legal chatbot have?
At minimum it should say it's an automated assistant, that it isn't a lawyer and can't give legal advice, that chatting doesn't create an attorney-client relationship, and that visitors shouldn't share confidential details. Put that in the greeting and above the contact form. Have your own ethics counsel approve the exact wording for your state.
Is a chatbot the unauthorized practice of law?
A chatbot that only gives information about your firm is very different from one that advises people on their legal problems. The Oregon opinion warns that a chatbot giving legal advice could implicate the lawyer in the unauthorized practice of law. Keep your bot to firm information and intake, and have it refuse anything about a visitor's own situation.