A divorce attorney’s job is no cakewalk. But even when things get fraught and desperate in family court, we’ll just point out the obvious: using AI to make up case law in your favor is not the answer, folks.
The Divorce Case That Started It All
The underlying case, Shahid v. Esaam, centered on a divorce in Georgia between wife Nimat Shahid and husband Sufyan Esaam. After the couple separated in July 2021, Shahid moved to Texas. In April 2022, Esaam filed for divorce in Georgia, represented by attorney Diana Lynch. Because he claimed not to know Shahid’s whereabouts, he served her by publication, a method of legal notice used when a party cannot be located after diligent efforts. In July 2022, the Georgia trial court granted a final judgment and decree of divorce based on this service by publication.
A year later, Shahid returned to court, represented by attorney Vic Brown, with a verified petition to reopen the case and set aside the final judgment. She argued that Esaam had not exercised the reasonable diligence required by Georgia law before resorting to service by publication, and therefore the court lacked jurisdiction over her. Shahid contended that because she could have been located with proper effort, the divorce decree was invalid.
The trial court held a hearing on Shahid’s petition. The core dispute was whether the divorce had been properly granted based on the method of service, and whether Shahid had been deprived of her right to contest the divorce due to improper procedure. Shahid argued that the trial court erred in denying her petition because her husband had not demonstrated due diligence, as required by OCGA § 9-11-4 (f), before using service by publication. The court ultimately denied Shahid’s request to reopen the case. In its order, the court cited two cases as legal justification for its decision.
But lucky for Shahid, her attorney had a better eye than the court staff: if nothing else, Brown realized that Lynch was cheating the system by making up cases.
When Fake Cases Come to Light
Brown appealed for his client. He pointed out that the trial court's reliance on two fictitious cases. The order was consequently “void on its face” due to reliance on non-existent law.
It was only during the appellate proceedings that the extent of the problem became clear: not only had the trial court’s order relied on fake case law, but Lynch had also submitted briefs containing numerous additional bogus or irrelevant citations apparently generated by artificial intelligence. The court noted that out of fifteen total citations in the Appellee’s Brief, eleven were to non-existent cases, and the remaining four were to real cases that were irrelevant to the arguments made.
Lynch didn’t do herself any more favors with her response when these claims surfaced. Rather than engaging directly with the core argument raised by Shahid’s attorney (that the trial court’s order was void because it relied on non-existent, “hallucinated” case law), Lynch sidestepped the issue entirely. Her brief did not attempt to explain or justify the reliance on fictitious cases in the lower court’s order. This omission was striking, given that Brown had expressly flagged the problem in her filings and made it central to her appeal.
Instead, Lynch’s appellate brief attempted to support her client’s position by citing four cases. But here, too, the problems multiplied. Two of the four cases she cited simply did not exist at all. These were not obscure or hard-to-find precedents; they were entirely fictitious, with citations that led nowhere in the official Georgia or regional reporters. The other two cases were real, but their content was irrelevant to the legal points for which Lynch cited them. She misrepresented their holdings or contexts, using them to support arguments they did not address.
Lynch’s missteps did not end with her use of questionable case law. In a further departure from standard appellate practice, she included a request for attorney’s fees for the costs incurred in responding to Shahid’s appeal. The court pointed out that not only was this request procedurally improper (since such motions must be filed separately rather than embedded in a brief), but Lynch attempted to justify her request by citing yet another fictitious case. She referenced “Johnson v. Johnson, 285 Ga. 408 (2009),” claiming it supported the award of attorney’s fees in divorce appeals. But the appellate court could find no such case by that name or citation. On top of that, the purported holding of Johnson did not reflect the actual state of Georgia law. In fact, Georgia courts have long held that the statute Lynch invoked, OCGA § 9-15-14, does not authorize the recovery of attorney’s fees for appellate proceedings.
The Cost of Cutting Corners
The court noted that the pattern of fictitious citations was consistent with the use of generative artificial intelligence, which is known to “hallucinate” cases that look convincingly real. And this was not a “no harm, no foul” situation. The court highlighted that making up legal authorities not only deprived the opposing party of a fair opportunity to respond but also wasted judicial resources and undermined the integrity of the appellate process. Perhaps the scariest part of all of this was the fact that Lynch got away with making up cases (for some time, anyway) and even won legal arguments with them.
Ultimately, Lynch’s approach resulted in the imposition of a $2,500 penalty for filing a frivolous motion. We wish we could say she’ll become a cautionary tale of the dangers of relying on generative AI for legal research, but let’s face it. She wasn’t the first attorney to do so, and she won’t be the last.
Related Resources:
- Haven't Started Using AI Yet? You Aren't as Behind as You Might Think. (FindLaw's Practice of Law)
- Alabama's Million-Dollar Prison Defense Team Caught Using Fake Citations (FindLaw's Practice of Law)
- Lindell Case Indicates Artificial Intelligence and the Legal System Are Still a Work in Progress (FindLaw's Practice of Law)
