Last week, we introduced Agentic AI and why solo attorneys and small firms should pay attention to it. This week, we’re going one level deeper, not into what AI can do for your firm, but into what it cannot do on your behalf, specifically on your website.
Your law firm website is not just a digital brochure. It is, in most states, a form of attorney advertising. Every word on it – whether written by you, a copywriter, or an AI tool – is subject to your state bar’s rules. Every response your website chatbot sends to a prospective client at 11pm is a communication made on your behalf. Every automated follow-up email triggered by a contact form is a message your firm sent.
That is the context in which AI on your law firm website has to operate. And it is the context many AI marketing tools were not specifically built to understand.
This post maps the lines that AI on your website cannot cross, and what it should do instead.
Your Website Has Three Jobs. AI Can Help With Two of Them.
Before drawing the lines, it helps to understand what your website is actually supposed to do.
At its core, a law firm website has three jobs:
- Attract the right visitors
- Build enough trust that they make contact
- Convert that contact into a consultation
AI tools can genuinely help with the first two. They can help you produce more content, more consistently. They can help your site show up in more searches. They can power a chatbot that responds to visitors at any hour. Those things all contribute to attracting visitors and building trust.
The third job is where things get complicated. Because that conversion moment is also the moment when the prospective client is most likely to share sensitive information, ask a specific legal question, or look to your website for reassurance about their situation.
That is exactly when AI is most capable of crossing an ethical line. And it is exactly when most attorneys are not watching.
The Line No One Told Your Chatbot About
Let’s start with the scenario that is already playing out on law firm websites.
A woman is on your personal injury firm’s website at 10:45pm. She was in a car accident three days ago. She has not spoken to an attorney yet. She finds the chat widget in the corner of your screen and types: “I was rear-ended and my back has been killing me since. Do I have a case?”
Your AI chatbot, which is trained to be helpful, warm, and conversational, responds: “Based on what you’ve described, it sounds like you may have a strong claim. Rear-end collisions often result in significant compensation for injuries like yours. Would you like to schedule a consultation?”
She screenshots it. She tells her friend. She calls your office the next morning expecting you to tell her she has a strong case, because your website already did.
There is a word for what just happened: the unauthorized practice of law (UPL). And there is a second word for the screenshot: evidence.
The AI chatbot on your website did not understand the difference between being helpful and providing legal advice. It was designed to convert. It converted. And in doing so, it created a liability your vendor agreement almost certainly does not cover.
The harm rarely shows up as a lawsuit. That woman moves on, files nothing, tells no one officially, and so it might feel like a clean outcome. But she does tell her sister, her coworker, someone. “I went to that firm’s website and they said I had a strong case, but then the attorney said it was complicated.” That one exchange becomes the story people tell about your firm, but you never hear it.
And if she does hire you? She arrives expecting a different conversation than the one you’re about to have. The damage isn’t legal. It’s the quiet erosion of trust between what your website promises and what reality delivers – repeated, invisibly, one visitor at a time.
The rule is simple, even if the line is thin: AI on your website can describe what your firm does. It cannot tell a prospective client what their legal situation means, what their options are, or what they should do. The moment it does any of those things, it has crossed from marketing into the practice of law.
What Your Chatbot Should Do Instead
A well-configured chatbot on a firm’s website is still valuable. It can acknowledge what the visitor shared (“It sounds like you’ve been through something really stressful. We’re glad you reached out.”). It can explain what your firm does. It can capture contact information. It can route the visitor to schedule a consultation.
So, you want to look for chatbots that offer the following characteristics or functions:
- It is designed for legal or professional services specifically and not repurposed from retail or e-commerce
- You can control exactly what it will and will not say
- It captures and routes information but does not provide any legal advice or conclusions
- It gives you access to full transcripts
- It has a data processing agreement you can actually read
Not all chatbots are created equal, and most are not built with attorneys in mind. Before you add one to your site, make sure you have checked the list above.
Content Published by a Law Firm on a Website Is Attorney Advertising
Now let’s talk about the content AI writes for your website, like the practice area pages, the blog posts, the attorney bios, because this is where bar advertising rules become directly relevant.
Every state bar has rules governing attorney advertising. Most of those rules were written before AI-generated content existed. All of them apply to it now.
Here is where AI content tools most likely can get law firm websites in trouble.
Outcome Language and Implied Guarantees
AI content tools default to persuasive language. That is what they are designed to produce. The problem is that persuasive language in legal advertising often crosses into prohibited territory.
An AI-generated headline like “How We Win Maximum Compensation for Injured Clients” sounds compelling. It is also likely a problem in most jurisdictions, because it implies a guaranteed outcome. The fix is simple enough: “How We Fight for Maximum Compensation for Injured Clients”. Though subtle, this is legally meaningful.
The same issue appears on case results pages. AI tools love to frame case outcomes as evidence of what you will do for the next client. Most bars require a disclaimer that past results do not guarantee future outcomes. AI tools do not add that disclaimer automatically. It has to be part of your review process.
The “Specialist” and “Expert” Problem
AI tools frequently reach for credentialing language to make attorney bios sound authoritative. The result is bios that describe attorneys as “specialists” or “experts” in practice areas where those terms are legally restricted.
In most states, an attorney cannot be described as a “specialist” in a practice area unless they hold formal board certification in that area from an organization approved by the state bar. AI tools do not know this unless they are told, and so will use “specialist” because it is a persuasive word.
If your AI-generated bio describes you as a “workers’ compensation specialist” but you do not hold that certification, your website might be in violation of your state’s advertising rules. And that’s true regardless of how many workers’ comp cases you have handled.
The fix is straightforward: review any AI-generated content for credentialing language before it goes live, and replace prohibited terms with accurate ones. “Focuses her practice on workers’ compensation.” “Has represented hundreds of clients in workers’ comp claims.” Both are accurate, compliant, and still compelling.
Superlatives That Cannot Be Verified
AI-generated taglines and practice area headers frequently include superlatives: “the most trusted,” “the leading,” “the top-rated.” These phrases are almost always impermissible under attorney advertising rules unless they can be independently verified through a recognized ranking or rating organization. And even then, the rules for using those rankings vary by jurisdiction.
“Phoenix’s most trusted DUI defense attorney” is not a claim your AI tool can substantiate. Neither can you. If it appears on your website, it is a problem.
The better approach – and the one that holds up under scrutiny – is to use specific, verifiable differentiators:
- Years in practice
- Number of cases handled
- Specific credentials or certifications
- Recognition from verifiable sources, cited properly.
These differentiators build credibility without creating compliance risk.
When a Law Firm Website Intake Form Becomes a Confidentiality Risk
There is a third category of ethical risk on law firm websites that gets less attention than it deserves: what happens to the information prospective clients share through your contact forms and intake tools.
Here is the scenario most attorneys have not thought through carefully enough.
A prospective client fills out your intake form with the details of their situation, including the facts of their accident, the nature of their criminal charge, the specifics of their custody dispute. That information flows into a CRM or lead management tool. An AI system then uses that information to draft a personalized follow-up email, summarize the intake for your review, or categorize the lead.
The question you need to be able to answer: where does that data go, how is it stored, and does the AI tool processing it have contractual protections in place that are consistent with your ethical obligations under ABA Model Rule 1.6?
Many consumer-grade AI tools do not. Their terms of service may allow them to use input data to improve their models. Their data retention policies may not meet the standards your ethical rules require. And unlike a law firm employee, they are not bound by a duty of confidentiality.
This does not mean you cannot use AI tools to manage intake. It means you need to know what you are agreeing to before you connect a prospective client’s information to any AI system. A data processing agreement (DPA) that explicitly prohibits use of your data for model training is a baseline requirement, not a nice-to-have.
The Invisible Drift Problem
Here is the thing about AI on your law firm website that most attorneys do not anticipate: even when it is set up correctly, it can quietly fall out of compliance – and not because the AI went rogue, but because your practice moved on and your AI configuration didn’t.
This is less about technology changing on its own and more about a gap that opens over time. Your firm added a new practice area. Your state bar issued updated guidance on AI-generated advertising. You stopped handling a certain case type. The AI on your website doesn’t know any of that. It keeps responding as if nothing has changed.
The more concrete version of this risk is vendor-side model updates. When an AI tool’s underlying model is updated, which happens regularly and often without user notification, its default behavior can shift. A chatbot configured to deflect legal questions conservatively may respond more confidently after an update. You didn’t change a setting. The system just changed.
The result in both cases is the same: a website that was compliant when you launched it but quietly becomes out of compliance six months later, with no single obvious moment where something went wrong.
Quarterly review is the practical fix, and should include:
- Reviewing chatbot transcripts
- Checking that AI-assisted content still reflects your current practice
- Comparing your website language against your state bar’s advertising rules
The quarterly review does not have to be exhaustive, but it should be consistent.
Three Questions to Ask Before Any AI Goes Live on Your Law Firm Website
If you take nothing else from this post, take these three questions. Run every AI-powered feature on your website through them before it goes live.
- Could a reasonable person read or receive this and believe they just received legal advice?
If the answer is yes, meaning a chatbot response, an automated email, or a piece of AI-generated content could reasonably be interpreted as a legal conclusion applied to someone’s specific situation, it has crossed the practice-of-law line. Fix it before it goes live.
- Does this content comply with my state bar’s advertising rules?
Review AI-generated content specifically for outcome language, credentialing terms, superlatives, and testimonial use. These are the four categories where AI tools most consistently produce content that violates bar rules. If you are unsure about a specific claim, the default position is to remove it.
- Does this AI tool have appropriate protections in place for the information prospective clients share through my website?
If you are connecting intake data, form submissions, or chatbot conversations to an AI tool, you need a data processing agreement. You need to know where the data is stored and for how long. You need to be able to tell a prospective client that the information they share through your website is handled with the same care your ethical rules require.
What Ethical Red Lines Look Like in Practice
Every practice area carries its own version of risks. Some are more exposed on the chatbot side, others in their website content, and others in how they handle intake data. Here is how it breaks down across the most common practice areas.
| Practice Area | Chatbot / Intake Risk | Website Content Risk | Data Confidentiality Risk |
| Personal Injury | AI implies victim has a strong case or that the statute of limitations “probably hasn’t run” | Settlement amounts posted without required disclaimers; outcome language like “we win maximum compensation” | Accident details, medical information, and witness facts shared through intake forms |
| Criminal Defense | AI advises prospect not to speak to police based on facts they’ve shared | Bio calls attorney a “top DUI specialist” without board certification; language implying charges will be dismissed | Arrest records, prior history, and case facts entered into unvetted AI tools |
| Family Law | AI tells a prospect courts “typically favor” an outcome that applies to their situation | Language promising to “win” custody; testimonials used without jurisdiction-specific disclaimers | Children’s names, school details, parenting schedules, and financial disclosures |
| Immigration | AI tells a visa holder they “should be fine” based on a brief description of their status | Content implying guaranteed outcomes on visa approvals or removal proceedings | Visa status, country of origin, documentation history — among the most sensitive intake data possible |
| Estate Planning | AI characterizes whether a will or trust “sounds valid” based on what a visitor describes | Outcome-implying language around asset protection or tax savings | Draft wills, trusts, and financial details uploaded to AI tools for editing |
| Workers’ Compensation | AI tells an injured worker they “have a strong claim” or advises them to file before seeing a doctor | “Specialist” label used without board certification; percentage-based recovery language without caveats | Employer details, injury circumstances, and medical history shared through automated intake |
| Business / Corporate | AI advises a prospect on whether a contract clause “looks enforceable” | Superlatives like “leading business attorney” without verifiable source | M&A documents, contracts, and financial details summarized through AI tools for marketing content |
The Standard Has Not Changed. Only the Temptation Has.
The ethical rules that govern your website were not written with AI in mind. They apply to it completely.
The attorney-client relationship. The duty of confidentiality. The prohibition on misleading advertising. These are not bureaucratic obstacles to efficient marketing. They are the reason clients trust attorneys with the most difficult or important moments of their lives. They are the reason your website carries more responsibility than most marketing assets do. And they are also the reason why the Model Rules of Professional Conduct were created, which was well before digital marketing was ever a thing.
AI does not change that standard. It raises the stakes for ignoring it.
The firms that will use AI most effectively on their websites are not the ones that automate the most. They are the ones that automate wisely with clear lines they will not cross, systems designed to enforce those lines, and an attorney who stays in the driver’s seat.
That is what good legal marketing has always required. AI just makes the question more urgent.
