Why ‘We Only Use AI For Marketing’ Is Still A Bar Complaint Waiting To Happen

A legal content agency owner warns that law firms using AI for marketing materials face bar complaints under ABA Model Rule 7.1, which bars false or misleading communications about a lawyer's services with no carve-out for websites, bios, or blog posts. The warning follows a pattern of AI-related sanctions: a New York lawyer was sanctioned in Mata v. Avianca in 2023 after ChatGPT confirmed six fabricated cases, a California appeals court imposed a $10,000 sanction in 2025 after finding 21 of 23 case quotations were fabricated, and by early 2026 a federal appeals court was charging attorneys $15,000 apiece, noting smaller fines had been inadequate.

Every lawyer who reads legal industry news knows what happens when AI errors show up in a court filing. A New York lawyer asked ChatGPT to confirm that six cases were real. It said yes. All six were fake, with invented docket numbers and quotations to match. That was Mata v. Avianca https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1:2022cv01461/575368/54/ in 2023, and Judge Castel sanctioned the lawyers over what he called bogus decisions with bogus quotes and bogus citations. The fines have only climbed since. Willkie Elevates Legal Work with Lexis+ with Protégé https://abovethelaw.com/2026/09/willkie-elevates-legal-work-with-lexis-with-protege/ Willkie AI and Innovation leader Todd Friedlich spoke to LexisNexis about firm’s thoughtful approach to legal AI A California appeals court hit a lawyer with a $10,000 sanction https://apnews.com/article/artificial-intelligence-general-news-california-courts-854e31420843daddfee622002d49338b in 2025 after finding that 21 of his 23 case quotations were fabricated, and published the decision as a warning. By early 2026, a federal appeals court was charging attorneys $15,000 apiece https://www.reuters.com/legal/litigation/us-appeals-court-fines-lawyers-30000-latest-ai-related-sanction-2026-03-16/ and noting that smaller fines had plainly been inadequate. So firms learned the lesson. Or they think they did. Here’s the logic I hear all the time: AI is dangerous in briefs because a judge checks your work, but marketing is a safe place to experiment. Run it on the website. Let it draft the bios. Spin up some practice area pages. If something is a little off, who’s going to sanction you over a web page? That assumption is the actual risk. And after 13 years running a legal content agency, I can tell you it does not hold up. PI Firms Don’t Need More Tools. They Need Fewer. https://abovethelaw.com/2026/09/pi-firms-dont-need-more-tools-they-need-fewer/ Every new tool a PI firm buys solves one problem and connects nothing. The real fix isn't another purchase — it's keeping the whole case in one place, from accident to resolution. Here's how. The Rule Does Not Care Whether It’s a Brief or a Bio ABA Model Rule 7.1 https://www.americanbar.org/groups/professional responsibility/publications/model rules of professional conduct/rule 7 1 communication concerning a lawyer s services/ is two sentences. A lawyer shall not make a false or misleading communication about the lawyer or the lawyer’s services. A communication is false or misleading if it contains a material misrepresentation of fact or law, or omits a fact necessary to make the statement, considered as a whole, not materially misleading. The official comment kills the safe-sandbox theory in one line. The rule governs all communications about a lawyer’s services, including advertising, and whatever means you use to make those services known, the statements have to be truthful. A website is a means. A bio is a means. A blog post with the firm’s name on it is a means. There’s no carve-out for the marketing department, because at most firms there is no marketing department. There’s just the lawyer whose name is on the page. Read the rest of that comment and it gets uncomfortable. A communication that truthfully reports a lawyer’s results can still be misleading if it leads a reasonable person to expect the same outcome in their own case. Unsubstantiated comparisons are flagged. So is any claim implying a credential or certification the lawyer doesn’t hold, which Rule 7.2 hits directly. Now picture the prompt a busy firm actually types: write a practice area page that highlights our track record and sets us apart from other firms in town. That prompt asks for exactly the claims the rule polices, and it hands the job to a tool that invents specifics for a living. What the Tool Does When You Ask It to Sell The data on legal AI is not comforting. A 2024 Stanford study https://hai.stanford.edu/news/ai-trial-legal-models-hallucinate-1-out-6-or-more-benchmarking-queries found general-purpose chatbots were wrong on legal questions more often than not, and even purpose-built legal research tools were wrong more than 17% of the time, with one major product hallucinating on more than a third of queries. Those numbers come from legal research, not marketing copy, so treat them as a floor. But the failure is the same one that matters here. The tool produces confident, polished, specific claims that aren’t true. In a legal brief, that looks like a fake citation. In a marketing page, it looks like a $2.3 million verdict the firm never won, a board certification the attorney never earned, or a Super Lawyers listing that doesn’t exist. The format is convincing either way. The only difference is what catches it. The Verification Gap Is the Whole Problem This is where the safe-sandbox theory has it backwards. Litigation isn’t the setting with the least error correction. It’s the setting with the most. Opposing counsel is paid to find your mistakes. A judge reads the pleadings. A motion to show cause drags the problem onto the record. Every AI sanction case became a case because someone went looking for a citation and couldn’t find it. A marketing page has none of that. No opposing counsel reviews your homepage. No judge reads your bio. A fabricated settlement figure can sit on a practice area page for months, indexed by Google, seen by every prospective client who lands there, with nothing in the process built to catch it. The first person to notice may be a competitor, a reporter, or the bar investigator opening a grievance. “Harder to Unwind” Is Not a Figure of Speech A bad brief gets fixed. You file an amended version, you tell the court and opposing counsel, and the audience for the mistake is small and the loop closes. A misleading web page doesn’t work that way. Bar discipline records are public and they last. Florida https://www.floridabar.org/public/acap/instructions/ posts attorney discipline documents with a searchable history measured in years. California https://www.statebarcourt.ca.gov/public-records-information makes discipline dockets available online by name, and in 2026 the state Supreme Court rejected a proposal https://www.marshalldennehey.com/lawyers-professional-liability/thought-leadership/california-supreme-court-rejects-automatic-expungement-of-attorney-disciplinary-records to auto-expunge those records after eight years. A grievance over a marketing claim becomes part of your permanent public profile no matter how it is resolved. It’s the first thing a prospective client sees when they search your name. The page itself is harder to erase than people assume. The Wayback Machine captures website content, and federal courts have accepted https://www.druganddevicelawblog.com/2021/07/its-back-its-wayback-its-away-wayback-its-admissible.html authenticated archive snapshots as proof of what a site used to say. Deleting the false claim doesn’t delete the evidence that it ran. And bar discipline isn’t the only exposure. Every state has an unfair and deceptive practices statute that reaches false advertising of professional services, enforced by attorneys general with their own penalties. The FTC has already moved here, settling with an AI legal services company https://www.ftc.gov/news-events/news/press-releases/2024/09/ftc-announces-crackdown-deceptive-ai-claims-schemes for $193,000 over claims that its product worked like a licensed lawyer. One false page can pull a bar grievance and a consumer protection complaint at the same time, from different directions. The exposure doesn’t stop with the bar or the FTC. A current client who relied on inaccurate information on your website and suffered an adverse outcome may even have a malpractice claim against you. By then, you’re defending the same false claim from three different directions—regulatory, consumer protection, and private tort The Honest Caveat, and Why It Doesn’t Save You Let me be straight about the record. No U.S. attorney has yet been publicly disciplined for AI-generated marketing content specifically. If you want to argue the risk is theoretical, that’s a fair argument. It’s also a weak one. The enforcement machine is fully built and has been running for decades on marketing claims that had nothing to do with AI. State bars have disciplined lawyers for advertising office locations that didn’t exist, for claiming credentials they couldn’t back up, for ads implying the firm couldn’t lose. The rule is settled. The process is active. The only new variable is a tool that manufactures the exact kind of false specific claim the rule was written to stop, at a volume no human copywriter could match. One AI session can fill an entire site with fabricated detail in minutes. The cases haven’t shown up yet because the practice is new, not because the conduct is safe. That’s the gap between “this hasn’t happened” and “this won’t happen,” and it’s closing. There’s a fair counterargument that the duty of competence already covers this. A careful lawyer reviews the output before it goes live and catches the fabrication. True. That’s also exactly what the sanctioned litigators believed they were doing. The ABA’s 2024 guidance on generative AI https://www.americanbar.org/news/abanews/aba-news-archives/2024/07/aba-issues-first-ethics-guidance-ai-tools/ is clear that the duty requires independent verification, not asking the AI whether it told the truth. If lawyers miss AI fabrications in a brief a judge is about to read line by line, betting they’ll catch them in marketing copy no one is checking is not a bet I’d make. The Actual Takeaway None of this is an argument against using AI in your marketing. We use it every day, and we’d be doing clients a disservice if we didn’t. It’s an argument against treating marketing as the place where the rules get loose. The standard on a practice area page is the same standard as on a brief, because Rule 7.1 says so in plain words. Verify the specifics, the results, the credentials, the awards, before they publish, with the same independent check you’d never skip in a filing. The sandbox was never safe. It just had no one watching it. That’s a different thing, and a worse one. David Arato is the founder of Lexicon Legal Content , a content marketing firm that creates SEO-optimized and consumer-focused content for law firms and digital marketing agencies throughout North America. He advises law firms on where AI belongs in their marketing content and where it does not, building the review gates that keep AI-assisted work accurate and bar-compliant. Get in touch on LinkedIn or email \ email protected\ https://abovethelaw.com/cdn-cgi/l/email-protection .