AI for Personal Injury Lawyers: Where It Helps at Intake and Where It Must Not Advise

Two kinds of AI in a personal injury firm

Most writing about AI for personal injury lawyers treats it as one thing. It is two.

The first kind works inside the firm. It reads medical records and builds a chronology, drafts the first version of a demand letter, summarizes a deposition or sorts a discovery production. Nobody outside the firm sees its work until a lawyer or paralegal has checked it. If it gets something wrong, the error is caught at the desk or it becomes the lawyer’s error.

The second kind works at the front door. It answers the person who reaches your website at 11 p.m., two days after a crash, with a sore neck and a voicemail from an adjuster. It speaks for the firm before any lawyer has read a word. Its mistakes reach the client directly.

The professional rules apply to both, but they bite differently. The rest of this guide is mostly about the second kind, because that is where a personal injury firm can do the most good and the most harm in a single conversation.

The back office: useful, with a lawyer in the loop

Back-office tools can save real hours on work that personal injury firms do over and over: turning hundreds of pages of medical records into a timeline, flagging gaps in treatment, drafting the routine parts of a demand, and summarizing transcripts. The American Bar Association’s Formal Opinion 512 on generative AI (July 2024) sets the ground rules, and they are familiar ones:

• Competence. You do not need to be a technologist, but you need a reasonable understanding of what the tool can and cannot do, and you must check its output before relying on it.
• Confidentiality. Read the tool’s terms. Medical records and settlement figures are exactly the information Rule 1.6 protects, and the opinion says informed client consent may be needed before confidential information goes into a tool that learns from what it is given.
• Supervision. Firm policies, training and oversight under Rules 5.1 and 5.3, including for outside vendors.
• Fees. Bill for the time actually spent, including reviewing the output, not for the time the tool saved.

None of this is new law. It is the old law applied to a new kind of assistant. The front door is where the details change.

The front door: what an injured person needs first

Someone who has just been hurt in a crash, a fall or a dog bite does not need legal analysis from a website. In the first conversation they need four things:

• A fast answer. They are often contacting more than one firm, and the first one to respond well usually gets the call back.
• The truth about who they are talking to. If it is software, it should say so at the start.
• No advice. They should leave the conversation without having been told what their case is worth or what to do about it, because nobody who could responsibly say has heard the facts yet.
• A clear next step. They should know their details are going to a person at the firm who will follow up.

Everything an intake tool does should serve one of those four.

Where AI intake helps a personal injury firm

A well-built intake assistant does work that is hard to staff around the clock:

• It answers at any hour. Accidents do not happen during office hours, and neither does the moment someone decides to look for a lawyer.
• It asks the same careful questions every time. What kind of accident, roughly when, and where it happened, including the state, because your service area depends on it.
• It asks whether an insurance company has been in touch. If an adjuster has already called, the lawyer will want to know that before the injured person says anything more to the insurer.
• It asks about pressure. Is anyone asking them to sign or accept something? Is there a deadline they have been told about? Those answers decide which inquiry a lawyer calls back first.
• It collects the other party’s name. A conflict check needs the name of the other driver, the store or the company, before anyone discusses the facts.
• It holds the conversation in the person’s language. An injured person who writes in Spanish or Haitian Creole should be able to finish the conversation in that language, while the firm receives the record in English. We explain why that matters in our guide to bilingual answering services.
• It reads the details back. A mistyped phone number is a lost client. The person should confirm their contact details before anything is sent.
• It stays out of medical detail. It can ask whether someone was hurt. It should not ask for diagnoses, medications or treatment records, which the lawyer will gather properly later.

If you want to see the questions laid out on paper, our free personal injury intake form template follows the same order.

Five questions an intake tool must not answer

Every injured person asks some version of these. Each one sounds factual. Each one is a legal judgment about that person’s situation, and an answer from software is legal advice given in your firm’s name.

“Do I have a case?” Whether there is a claim depends on facts the intake tool has not tested and on law it should not be applying. The honest answer is that an attorney will review what they have shared. A tool that says “it sounds like you have a strong case” has made a promise the firm may not keep.

“What is my case worth?” Settlement calculators and value ranges are everywhere on the web, and some intake tools offer them. Do not let yours. A number given at first contact anchors the client’s expectations before anyone has seen a medical record, and it can read as a misleading statement about results the firm can obtain.

“Was it my fault?” In many states this question can decide the whole claim. Some states reduce what an injured person recovers by their share of the fault; many bar recovery entirely once that share passes a set threshold; a few bar it for almost any fault at all. Which rule applies, and where one person’s facts fall under it, is exactly the analysis a lawyer is for. An intake tool should take down what happened in the person’s own words and stop there.

“Should I talk to the adjuster, give a recorded statement or take the offer?” This is advice in its purest form. The tool can record that an insurer has called and that an offer is on the table, and mark the inquiry as urgent. It should not say yes or no.

“How long do I have?” Deadlines are real, and in many states some are short. Every state limits how long an injured person has to bring a claim, the limits differ by state and by kind of claim, and legislatures change them. A claim against a government body can require written notice much sooner than the general deadline, and some no-fault insurance benefits depend on getting medical care promptly after the accident. But telling one person which deadline applies to them, and when it runs out, is applying the law to their facts. What the intake tool should do instead is collect the date of the accident and anything time-sensitive, and put those at the top of the record so a lawyer calls back fast. The deadline is the reason for speed at intake, not something for the tool to explain.

A sixth question comes up too: “Should I see a doctor?” An intake tool is not a medical adviser either. If someone describes an emergency, the right answer is to call 911. Otherwise, the medical questions belong with medical professionals and the legal ones with the lawyer.

Disclosure, supervision and the bar

The rules for client-facing AI are now written down. The Florida Bar’s Ethics Opinion 24-1 (January 2024) says a chatbot that communicates with prospective clients must make clear that it is an AI program and not a lawyer or employee, and that the lawyer remains responsible if it gives misleading information. It also suggests screening questions, for example to stop the chatbot from going further with someone who already has a lawyer. The ABA’s Formal Opinion 512 flags the advertising rules for client-facing tools and points to Florida’s analysis.

Three practical consequences follow for a personal injury firm:

• Say it at the start. The first message should say that the visitor is talking to an AI assistant, that it is not an attorney, and that nothing in the conversation is legal advice.
• Supervise it under Rule 5.3. Know every sentence it can say, test it on hard cases before launch, and read what it collects.
• Collect only what you need. What a prospective client tells a firm is protected under Rule 1.18, and a firm that learns too much from someone it cannot represent can create problems for itself on the other side of the case. A short, structured intake is safer than a long free-form one.

We cover the full set of rules in our legal intake compliance guide.

Seven questions to ask an AI intake vendor

1. What does the first message say? Ask to see it word for word, in every language the tool speaks.
2. What happens when someone asks what their case is worth? Ask for the exact reply.
3. Does it ask whether an insurer has been in touch, and whether anyone is pressuring the person to sign or accept something?
4. Does it collect the other party’s name for a conflict check?
5. What does it do with medical detail? It should not ask for diagnoses or treatment, and it should not repeat them back.
6. Does the person confirm their details before anything is sent, and what exactly does the firm receive? Ask for a sample record.
7. What happens when someone describes an emergency, a threat or a minor child? The answer should be specific and written down.

The bottom line

AI can take hours of document work off a personal injury firm’s desk, and it can make sure the injured person who writes at midnight gets a careful answer instead of a contact form. The difference between those two jobs is who checks the work. In the back office, a lawyer checks it before anyone relies on it. At the front door, nobody does, so the tool has to be built to stop exactly where the lawyer’s job begins. For how intake fits together as a whole, start with our complete guide to legal intake.

Common questions

Can AI handle intake for a personal injury firm?
It can handle the first conversation: answering at any hour, saying it is an AI assistant, asking the intake questions, collecting contact details and the other party’s name, and sending the firm a structured record. It should not evaluate the case, estimate its value or advise the person, which remains the lawyer’s job.

Is it ethical for a personal injury firm to use an AI chatbot on its website?
Yes, if the firm follows the rules that apply to any assistant. The Florida Bar’s Ethics Opinion 24-1 requires a clear disclosure that the chatbot is AI and not a lawyer, keeps the lawyer responsible for what it says, and suggests screening questions. The firm must also supervise it under Rule 5.3 and protect what prospective clients share.

Should an intake chatbot tell someone what their case is worth?
No. A value given at first contact is a legal judgment made without the facts, it anchors the client’s expectations and it can be a misleading statement about results. The intake should record what happened and pass it to an attorney.

What should an AI intake tool ask after a car accident?
What kind of accident it was and roughly when, where it happened (including the state), whether an insurance company has been in touch, whether anyone is pressuring the person to sign or accept anything, the name of the other driver or company for the conflict check, and the best way to reach them. It should not ask for medical detail.

Can AI help with medical records and demand letters?
Yes, as a back-office tool. It can summarize records, build chronologies and draft first versions of demands, but a lawyer must review the output before relying on it, and the firm must check the tool’s terms before confidential records go into it, as the ABA’s Formal Opinion 512 explains.

Related: Legal Intake, Explained — our complete guide · Legal intake compliance guide · Free client intake form templates · AI receptionist for law firms · Bilingual answering service for law firms · all guides.

Ethics guidance cited from ABA Formal Opinion 512 (July 29, 2024) and Florida Bar Ethics Opinion 24-1 (January 19, 2024), both checked October 21, 2026. Fault rules and filing deadlines vary by state; nothing on this page is legal advice.