The Arizona Court of Appeals put lawyers and self-represented litigants alike on notice: If you let artificial intelligence write your legal brief and don’t check whether the cases it cites are real, you can be sanctioned for it — even if you never meant to mislead anyone.

“Submission of hallucinated case citations — intentional or unintentional — is unjustifiable. To accept a defense of lack of ill intent would legitimize reckless indifference to the truth and accuracy of court filings and tolerate some level of misrepresentation,” Judge Brian Furuya wrote. “This we will not do.”

Image

The ruling by the appellate court marks the first time that Arizona courts have published an opinion on the use of generative artificial intelligence — and the hallucinations that come with it — in legal filings.

The unanimous ruling, and the sanctions that it imposes, sprang out of a probate case. Gloria Acciavatti died in June 2023 at 93, leaving three adult children. When her daughter, Kim Dineen, sought appointment to administer the estate, she found there was nothing to administer because the assets were titled in her brother Jeffrey Kotchka’s name.

Dineen went to court in February 2024 to have the estate sorted out and herself named personal representative. Both of her siblings objected. After a hearing, a Maricopa County judge sided with Dineen and appointed her in July 2025.

Kotchka, representing himself, appealed. In his opening brief with the Court of Appeals, he made eight citations. The court found that six of them were deficient: Some pointed to wrong pages, some attributed rules or facts to cases that didn’t contain them, and two were cases that “simply do not exist.”

In a written explanation and at a court hearing focused on the false citations, Kotchka readily acknowledged that he relied on generative AI to write his brief, and he conceded that he had not used — or even seen — the materials available on the Court of Appeals’ website for self-represented litigants. Among those materials is a guide providing tips and warnings for using AI tools.

Image

And Kotchka testified that he didn’t use any publicly available research tools to verify the arguments and citations in his filing, nor did he consult a professional.

“Indeed, he admitted he did nothing to independently verify the citations. We appreciate his candor, but all litigants, whether appearing through lawyers or in their own person, must ensure their filings comply with all applicable rules and law,” Furuya wrote.

Reliance on generative AI doesn’t absolve parties in a case, whether they are licensed attorneys or laypeople representing themselves, from abiding by court rules, the appellate court ruled.

“Accordingly, citing a hallucinated case in a legal filing is sanctionable conduct,” Furuya wrote.

The three-judge panel noted that court rules allow for more severe sanctions, including contempt charges and monetary fines. But because Kotchka claimed he didn’t intend to deceive the court and isn’t a trained legal professional, it opted to narrowly sanction him by requiring him to pay the portion of Dineen’s attorney’s fees traceable to dealing with the fake citations. How much that is depends on what documentation Dineen submits to the court.

Kotchka’s appeal was also rejected because there were no supportive legal authorities for his claims once the inaccurate and false citations were stricken.

Although it set a framework for Arizona courts to sanction parties in the future for AI hallucinations, the Court of Appeals was careful to say the problem isn’t the technology. Generative AI can check grammar, tighten prose and summarize long documents, Furuya wrote, and the panel said it “recognize[s] the promise” such tools hold for people who can’t afford a lawyer, “provided the person signing the brief checks to make sure all the cases they cite are real and accurately cited.”