Each lawyer who reads authorized business information is aware of what occurs when AI errors present up in a court docket submitting. A New York lawyer requested ChatGPT to verify that six circumstances have been actual. It mentioned sure. All six have been pretend, with invented docket numbers and quotations to match.
That was Mata v. Avianca in 2023, and Decide Castel sanctioned the legal professionals over what he referred to as bogus choices with bogus quotes and bogus citations. The fines have solely climbed since.
A California appeals court docket hit a lawyer with a $10,000 sanction in 2025 after discovering that 21 of his 23 case quotations have been fabricated, and printed the choice as a warning. By early 2026, a federal appeals court docket was charging attorneys $15,000 apiece and noting that smaller fines had plainly been insufficient.
So companies discovered the lesson. Or they suppose they did.
Right here’s the logic I hear on a regular basis: AI is harmful in briefs as a result of a decide checks your work, however advertising and marketing is a protected place to experiment. Run it on the web site. Let it draft the bios. Spin up some apply space pages. If one thing is a bit of off, who’s going to sanction you over an internet web page?
That assumption is the precise threat. And after 13 years working a authorized content material company, I can inform you it doesn’t maintain up.
The Rule Does Not Care Whether or not It’s a Temporary or a Bio
ABA Model Rule 7.1 is 2 sentences.
A lawyer shall not make a false or deceptive communication concerning the lawyer or the lawyer’s providers. A communication is fake or deceptive if it accommodates a fabric misrepresentation of truth or legislation, or omits a truth essential to make the assertion, thought of as a complete, not materially deceptive.
The official remark kills the safe-sandbox idea in a single line. The rule governs all communications a couple of lawyer’s providers, together with promoting, and no matter means you employ to make these providers recognized, the statements should be truthful.
A web site is a method. A bio is a method. A weblog publish with the agency’s title on it’s a means. There’s no carve-out for the advertising and marketing division, as a result of at most companies there isn’t any advertising and marketing division. There’s simply the lawyer whose title is on the web page.
Learn the remainder of that remark and it will get uncomfortable. A communication that honestly stories a lawyer’s outcomes can nonetheless be deceptive if it leads an affordable particular person to anticipate the identical end result in their very own case. Unsubstantiated comparisons are flagged. So is any declare implying a credential or certification the lawyer doesn’t maintain, which Rule 7.2 hits immediately.
Now image the immediate a busy agency really varieties: write a apply space web page that highlights our observe document and units us other than different companies on the town. That immediate asks for precisely the claims the rule polices, and it palms the job to a software that invents specifics for a dwelling.
What the Instrument Does When You Ask It to Promote
The info on authorized AI shouldn’t be comforting. A 2024 Stanford study discovered general-purpose chatbots have been incorrect on authorized questions most of the time, and even purpose-built authorized analysis instruments have been incorrect greater than 17% of the time, with one main product hallucinating on greater than a 3rd of queries.
These numbers come from authorized analysis, not advertising and marketing copy, so deal with them as a flooring. However the failure is identical one which issues right here. The software produces assured, polished, particular claims that aren’t true.
In a authorized temporary, that appears like a pretend quotation. In a advertising and marketing web page, it seems like a $2.3 million verdict the agency by no means gained, a board certification the lawyer by no means earned, or a Tremendous Attorneys itemizing that doesn’t exist. The format is convincing both means. The one distinction is what catches it.
The Verification Hole Is the Complete Drawback
That is the place the safe-sandbox idea has it backwards. Litigation isn’t the setting with the least error correction. It’s the setting with probably the most.
Opposing counsel is paid to seek out your errors. A decide reads the pleadings. A movement to indicate trigger drags the issue onto the document. Each AI sanction case grew to become a case as a result of somebody went in search of a quotation and couldn’t discover it.
A advertising and marketing web page has none of that. No opposing counsel critiques your homepage. No decide reads your bio. A fabricated settlement determine can sit on a apply space web page for months, listed by Google, seen by each potential shopper who lands there, with nothing within the course of constructed to catch it. The primary particular person to note could also be a competitor, a reporter, or the bar investigator opening a grievance.
“Tougher to Unwind” Is Not a Determine of Speech
A foul temporary will get fastened. You file an amended model, you inform the court docket and opposing counsel, and the viewers for the error is small and the loop closes. A deceptive internet web page doesn’t work that means.
Bar self-discipline information are public and so they final. Florida posts lawyer self-discipline paperwork with a searchable historical past measured in years. California makes self-discipline dockets accessible on-line by title, and in 2026 the state Supreme Court docket rejected a proposal to auto-expunge these information after eight years.
A grievance over a advertising and marketing declare turns into a part of your everlasting public profile regardless of how it’s resolved. It’s the very first thing a potential shopper sees after they search your title.
The web page itself is tougher to erase than individuals assume. The Wayback Machine captures web site content material, and federal courts have accepted authenticated archive snapshots as proof of what a website used to say. Deleting the false declare doesn’t delete the proof that it ran.
And bar self-discipline isn’t the one publicity. Each state has an unfair and misleading practices statute that reaches false promoting {of professional} providers, enforced by attorneys basic with their very own penalties. The FTC has already moved right here, settling with an AI legal services company for $193,000 over claims that its product labored like a licensed lawyer. One false web page can pull a bar grievance and a shopper safety grievance on the similar time, from totally different instructions.
The publicity doesn’t cease with the bar or the FTC. A present shopper who relied on inaccurate info in your web site and suffered an adversarial end result might actually have a malpractice declare in opposition to you. By then, you’re defending the identical false declare from three totally different instructions—regulatory, shopper safety, and personal tort
The Sincere Caveat, and Why It Doesn’t Save You
Let me be straight concerning the document. No U.S. lawyer has but been publicly disciplined for AI-generated advertising and marketing content material particularly. If you wish to argue the chance is theoretical, that’s a good argument. It’s additionally a weak one.
The enforcement machine is absolutely constructed and has been working for many years on advertising and marketing claims that had nothing to do with AI. State bars have disciplined legal professionals for promoting workplace places that didn’t exist, for claiming credentials they couldn’t again up, for adverts implying the agency couldn’t lose.
The rule is settled. The method is energetic. The one new variable is a software that manufactures the precise sort of false particular declare the rule was written to cease, at a quantity no human copywriter might match. One AI session can fill a whole website with fabricated element in minutes.
The circumstances haven’t proven up but as a result of the apply is new, not as a result of the conduct is protected. That’s the hole between “this hasn’t occurred” and “this gained’t occur,” and it’s closing.
There’s a good counterargument that the responsibility of competence already covers this. A cautious lawyer critiques the output earlier than it goes dwell and catches the fabrication. True. That’s additionally precisely what the sanctioned litigators believed they have been doing. The ABA’s 2024 guidance on generative AI is evident that the responsibility requires unbiased verification, not asking the AI whether or not it advised the reality.
If legal professionals miss AI fabrications in a short a decide is about to learn line by line, betting they’ll catch them in advertising and marketing copy nobody is checking shouldn’t be a wager I’d make.
The Precise Takeaway
None of that is an argument in opposition to utilizing AI in your advertising and marketing. We use it day-after-day, and we’d be doing shoppers a disservice if we didn’t. It’s an argument in opposition to treating advertising and marketing because the place the place the foundations get free.
The usual on a apply space web page is identical commonplace as on a short, as a result of Rule 7.1 says so in plain phrases. Confirm the specifics, the outcomes, the credentials, the awards, earlier than they publish, with the identical unbiased verify you’d by no means skip in a submitting.
The sandbox was by no means protected. It simply had nobody watching it. That’s a distinct factor, and a worse one.
David Arato is the founding father of Lexicon Legal Content, a content material advertising and marketing agency that creates Search engine optimisation-optimized and consumer-focused content material for legislation companies and digital advertising and marketing companies all through North America. He advises legislation companies on the place AI belongs of their advertising and marketing content material and the place it doesn’t, constructing the overview gates that maintain AI-assisted work correct and bar-compliant. Get in contact on LinkedIn or e-mail david@lexiconlegalcontent.com.
The publish Why ‘We Only Use AI For Marketing’ Is Still A Bar Complaint Waiting To Happen appeared first on Above the Law.