Your Law Firm’s Marketing Needs an AI Policy

Lawyers defending State Farm in a homeowner’s lawsuit were just caught using AI in their court case filings. 7 citations across 8 filings contained AI hallucinations, referencing cases that did not exist and quotes that were made up.

While this may be the latest instance of a blunder like this, it’s not the first time we’ve seen an attorney put their faith in the hands of AI in the courtroom.

Last year, California courts issued a $10,000 fine to an attorney who had an opening brief that had 21 fabricated quotes out of 23 used. It was later determined that the lawyer used ChatGPT to create the brief.

Additionally, a database tracking legal AI hallucinations in court filings has identified over 1,900 cases worldwide (more than 1,300 of those in the United States).

Outside of the courtroom, these same types of failures can happen quietly on the legal marketing side of things. There’s just no opposing counsel reading every line of your website or judge to catch it.

But the consequences can be equally harmful to your legal practice, which is why you need to clearly define an AI policy for your marketing materials.

The Legal Industry Has the Worst Hallucination Rates

There’s no doubt that AI is getting better seemingly by the day. But any reassuring statistics that you’ve seen about its accuracy can be thrown out the window when it comes to legal marketing.

Here’s what the latest data from Seekr says:

Reports show some models boasting a 2% hallucination rate. However, those are based on handing the model a clean document and asking for a summary. But if you ask AI models to produce facts without giving it a source, error rates jump anywhere from 16% to 33%.

It gets even worse in the legal industry, as legal queries have AI hallucination rates in the 69% to 88% range.

Legal-specific tools have narrowed the gap slightly. But they’re still far from reliable.

A preregistered study by Stanford and Yale researchers evaluated legal research products leveraging AI, and the results weren’t great:

ChatGPT-4 hallucinated in 43% of the legal searches tested.
Westlaw AI-Assisted Research hallucinated in roughly one-third of queries, and only returned accurate answers 42% of the time.
Lexis+ AI produced hallucinated or misleading answers in 17% of legal searches.
Despite the lower hallucination rate, Lexis+ AI only gave an accurate, grounded answers 65% of the time.

Over-reliance on these types of tools without proper oversight can get you into trouble, and that’s exactly what happened in the State Farm case I mentioned earlier.

The lawyer was using an AI tool that she thought was connected to Westlaw, and assumed an internal citation check was being conducted. That didn’t happen.

In a YMYL Category, There’s No Acceptable Threshold for Inaccuracies

Shifting this back to the marketing angle. It’s important for you to understand that Google classifies legal content as Your Money or Your Life (YMYL), a category with the highest possible stakes.

It covers topics affecting a person’s health, financial stability, safety, or wellbeing.

So the highest standards are applied to these types of searches, and E-E-A-T signals carry the most weight here. While we don’t have the same algorithm information for AI models, it’s clear that they only cite and recommend sources that are trustworthy.

Point being this: If your law firm website contains made-up cases, quotes that don’t exist, or inaccurate legal claims, it can crush your ability to rank in both traditional SEO and AI SEO standards.

Even if it’s not on your website, and you’re caught in one of these stories gaining attention in the courtrooms, that negative press is going to signal to Google’s algorithm and AI models that your firm is not trustworthy enough to source or refer.

Let’s put the rankings, traffic, and leads aside for a moment.

Think about the consequences that inaccurate information and a law firm’s website can have in the real world. Say you publish a blog post or landing page using AI that contains a deadline to file a claim in your state. But the date is off by six months and nobody fact-checks it.

A reader can now lose their opportunity to file because your firm’s name told them they had more time.

Where AI Belongs in Legal Marketing, and Where It Does Not

The most useful way to approach this isn’t “AI content” vs. “human content,” and that’s often where I see firms getting tripped up.

Instead, the decision on whether you should be using AI for a legal task depends on whether the answer already exists in the material you’re handing the tool or whether the AI has to produce something on its own.

That’s the distinction that correlates to the error rates mentioned above. And it’s specific enough for you to write in your AI legal marketing policy.

It’s typically ok to use AI to work from source materials that you provide, such as:

Summarizing case files or discovery documents for internal meetings
Turning a recorded attorney podcast interview into a blog outline or draft
Transcribing and organizing notes from a client intake
Restructuring case notes into inspiration for a new blog, landing page, video, or podcast
Repurposing a piece you already published into a newsletter or social media post
Drafting subject lines for emails
Editing content for clarity, flow, and consistency against your style guide

In these scenarios, you’re not asking AI to produce legal facts for you.

Where you can get into trouble by having your receptionist open ChatGPT and type, “write me a blog post about wrongful termination laws in Texas.”

Nobody gave the model anything here. And that’s when it’s most likely going to hallucinate facts or make up cases. Same goes for any request to explain laws, legal standards, deadlines, or case citations without producing a source document.

What Always Needs to Be Verified

Whether you have an in-house marketing team or using an agency, there should be a human check by a lawyer in your practice for any marketing materials produced.

Blogs, landing pages, FAQs, emails, podcasts, YouTube videos. All of it, regardless of how it was produced, if the content contains:

Case citations or quotes
Statutes of limitations
Filing deadlines
Any jurisdiction-specific legal statement
Attorney credentials
Past case results and settlement amounts
All statistics
Medical claims or scientific research

With these types of claims, you’re not only exposed to potential problems from an SEO standpoint. But you’re also facing state bar advertising rules that prohibit false or misleading communication.

AI is really good at making things sound right, even if it’s not true. That’s why it’s so risky to use in these situations without a human check.

How to Write Your Law Firm’s AI Marketing Policy

The first thing you need to do is start with an inventory of where you stand right now.

One of the biggest problems in the State Farm case was not that the attorney used AI. It was that they were using a tool that nobody had evaluated and used in a way that they thought was connected and verified to something legitimate.

Ask everyone in your firm what types of tools they’re using. Attorneys, paralegals, legal assistants, marketing coordinators, and any outside agency or freelancer that touches your content. Get a list and research all of them.

From there, the policy should include:

Approved tools with a statement that nothing else can be used.
A requirement that an attorney review anything of substance before it goes live.
Specific workflow steps that require verifications as a safeguard, with a named owner who is responsible for it.
Training on what these tools do badly.
Guidance on the types of materials that can be fed to AI, and prompts that produce accurate responses.

None of this requires you to “ban” AI.

It’s all about having established rules so mistakes don’t happen.

Final Thoughts

Even as AI continues to make major advancements, it still hallucinates and makes mistakes. And that’s unacceptable for legal marketing.

The stakes are higher in this space, and that’s why the consequences can be so severe. A $10,000 fine is a slap on the wrist when you compare it to the potential damage to your firm’s reputation.

So take this seriously before it’s too late. You may also want to audit any content that’s been produced over the last 6-12 months to see if there’s any factual errors or made up cases living on your website right now.