by John P. Kamin
Tuesday, January 20, 2026 | 0
If there’s anything lawyers have learned about artificial intelligence over the past few years, it’s that they shouldn’t trust it to cite cases.
John P. Kamin
In December 2025, a kindly neighbor judge asked if I had heard about the latest ruling imposing sanctions for miscitations. To be honest, I wasn’t too sure about this recent incident. Because there have been so many incidents in recent years.
For example, the California Applicants’ Bar Association wrote an article about one such case in September. And WorkCompAcademy wrote about a similar but different case in December. Long before these articles, a court in 2023 sanctioned a New York lawyer for using citations from as many as six fictitious cases, drawing the attention of the state bar association. To be clear, the judge I spoke to did not mention any of these cases. He was talking about a new case.
why?
Why is this happening? AI-powered engines can sometimes make things up out of thin air, a phenomenon known as “hallucinations.”
Lawyers in a hurry to prepare arguments can simply ask the AI chatbot for case citations and legal summaries, then “copy-paste” them into court briefs, petitions, and appellate briefs. (Copy-pasting is when someone sends a large block of copied and pasted text to a bulletin board or social media app. The text is often silly in nature.)
Judges and their clerks presiding over these attorneys’ cases should endeavor to scrutinize these citations. (Interestingly, these people are now in a similar position to high school and college teachers who have to decide whether the work at hand is real or just plagiarized garbage.)
If a judge can’t find those citations anywhere and turns around and asks lawyers to take those cases, those lawyers can’t produce those citations. Because they simply don’t exist.
Yay! It’s not a good place.
How to avoid this problem
Whether you’re using AI, Google, or your favorite law paper to do your legal research, there’s an easy way to get around this problem. Once you find the citations, pull out the specific case and actually read it.
LexisNexis should have most of these cases. If they’re too expensive or hard to find, many of these cases can also be found online.
When you read actual opinions, you may find nuances that support or refute your argument.
Finally, when I read the case cited by the opponents, I found that it gave the defendant a victory on that very issue. As a result, I was pleased to see opposing counsel make such an argument and was happy to point it out to the trial judge.
(The cited case featured multiple sensitive sub-issues on the same subject that the opposing attorney thought was in favor of the plaintiff. However, a closer reading showed that the defendant won on my sub-issues. The summary did not mention my sub-issues, so you would not know that without reading the case.)
Reading case law has other benefits as well. This gives you the opportunity to quote text from the opinion. There is no need to reinvent the wheel when a wise and enlightened previous judge/chairman/appellate judge said it better than anyone else years ago. And of course, after quoting it, you can also paraphrase it before or after it for dramatic effect.
If you cite previous opinions, we recommend that you cite them and provide attribution where attribution is required. When I was a journalist, I often used exact quotations when I wasn’t sure how to summarize a complex legal analysis. Because it’s often the most accurate method.
In summary, read the examples you cite, or at least make sure they exist and generally support the proposition you’re citing. There are countless ways to help you. If you don’t read it, you’re only missing out.
What not to do
One tactic that repeatedly fails when asked to explain AI hallucinations is to blame the AI chatbot. Blaming bots for work that lawyers are billing has failed in every jurisdiction where I’ve seen this fact pattern occur.
The judges consider that explanation merely lazy and excusable.
In such a scenario, you should choose the path that causes the least discomfort, even if it means eating some AI crow.
conclusion
Be aware that AI chatbots can hallucinate false quotes and get you into a ton of trouble. If these kinds of troubles come together, your reputation can be damaged.
Don’t fall into that trap. Please check and read the examples cited. You will gain more knowledge.
John P. Kamin is a workers’ compensation attorney and partner in Bradford & Barthel’s Woodland Hills office. He is a former legal editor at WorkCompCentral. This entry from Bradford & Barthel’s blog is published with permission.
