Sheerer v. Panas: California Court Extends AI Hallucination Warning to Pro Se Litigants | Advocate Prakhar

Case at a Glance

Background

Anna Sheerer and Thomas Panas divorced after a lengthy custody dispute over their two children. The case came to the Court of Appeal on Sheerer’s challenge to a September 2024 child support order that she argued failed to account for Panas’s substantial bonus income and restricted stock units (RSUs). Panas, representing himself, filed a respondent’s brief in opposition. On reviewing the brief, the court found it cited at least two non-existent cases — “In re Marriage of Thomson (1984) 157 Cal.App.3d 568, 578” and “Marriage of Mendlowitz (2019)” — and attributed quotations to real cases where no such words appear. Panas admitted in a declaration that he had used generative AI to write the brief.

The AI Issue

Panas’s AI-generated brief contained multiple layers of fabrication: non-existent cases (citations that simply do not appear in any legal database), fabricated quotations attributed to real cases (where the cited language is absent from the actual opinions), and authority cited for propositions entirely unrelated to the points for which they were cited. The court exercised its discretion under California Rules of Court, Rule 8.276(a)(4), to impose sanctions for unreasonable violations. It chose to disregard the entire brief rather than impose monetary sanctions, citing the best interests of the children at the center of the case. However, it published the AI hallucination section of its opinion as binding precedent.

What the Court Decided

• Panas’s respondent’s brief was disregarded entirely due to “extensive reliance on nonexistent legal authority.” [brief striking]
• Pro se litigants are held to the same standard as attorneys — the court explicitly refused to create a “separate track” for self-represented litigants using AI. [equal accountability]
• The child support order of September 2024 was reversed on the merits because the trial court failed to account for Panas’s bonus and RSU income under Family Code § 4058. [mandatory income inclusion]
• Monetary sanctions were declined because their imposition would harm the children’s financial interests. [best interests exception]
• The AI hallucination warning section was certified for PUBLICATION, creating binding precedent in California courts. [precedential force]

“Our warning to litigants is not merely an admonition to double-check citations and otherwise fastidiously comply with the Rules of Court; it is to be at all times truthful and to be responsible in crafting any written arguments presented in this Court.”

— Cal. Ct. App. First District, Div. 4, March 19, 2026

The India Angle

Indian Law Equivalent

Under the Hindu Marriage Act 1955 and the Special Marriage Act 1954, income disclosure in maintenance and alimony proceedings (Sections 24–25 HMA) requires accurate financial data. Bonus income and ESOPs are treated as part of gross income for maintenance calculation in India — see Bharat Hegde v. Sarla Hegde (Bombay HC). Any fabricated citation in maintenance-related pleadings could constitute contempt under Section 2(d) of the Contempt of Courts Act 1971.

Bar Council Rules

Rule 49 of the BCI Rules requires advocates to be scrupulously accurate in all statements of fact and law. The California court’s published warning — that using AI without verification is a breach of truthfulness obligations — maps directly onto Rule 49 and Rule 32 (prohibition on fraud or deceit). Indian family law practitioners increasingly use AI to generate maintenance calculation worksheets and draft orders; each citation in such documents must be independently verified.

Practical Advice for Indian Advocates

• Treat AI-generated maintenance calculation briefs as drafts requiring full citation verification — bonus income and ESOP calculations involve complex judicial interpretation that AI may misstate.
• In child support or maintenance cases, courts are especially sensitive to anything that undermines the children’s interests; fabricated citations in such proceedings may draw harsher judicial response than in commercial disputes.
• Build a two-step verification protocol: first, confirm the case exists; second, confirm the specific quoted language appears in the actual judgment, not just the headnote or AI summary.

Quick Takeaways

• California’s published ruling creates binding precedent that pro se AI use without verification is sanctionable.
• Courts can disregard an entire brief for AI hallucinations, regardless of any meritorious arguments it contains.
• Best interests of children can reduce the severity of sanctions but does not eliminate liability for the AI hallucination violation.

Deep Dive: California’s Published AI Warning and Its Ripple Effects

Sheerer v. Panas is one of the most consequential AI hallucination rulings of 2026 because it is published — meaning it carries binding precedential weight in California appellate courts and persuasive weight in courts nationwide. The court’s decision to certify only the AI hallucination sections for publication while keeping the child support merits analysis unpublished was a deliberate choice: California courts publish opinions that establish new legal principles or resolve conflicts, and the court plainly viewed the AI hallucination issue as meeting that threshold.

The California precedential landscape on AI hallucinations now includes both Noland v. Land of the Free, L.P. (2025) — which warned attorneys — and Sheerer v. Panas — which extends that warning to self-represented litigants. Together, these cases create a comprehensive framework: anyone who files a document in a California court, regardless of legal training, must independently verify every citation. The days of credibly claiming that one “didn’t know” about AI hallucination risks in California are effectively over.

The court’s reasoning on why pro se litigants cannot receive special treatment is worth examining closely. The opinion traces the history of California’s equal-standard doctrine back to Rappleyea v. Campbell (1994), which warned that a doctrine allowing exceptional treatment for self-represented parties would create “a quagmire in the trial courts.” The court then noted that this concern has “grown truer in an age where litigants can access AI tools that purport to do lawyering.” This framing is significant: AI tools have lowered the practical bar to filing complex legal documents, which means the equal standard becomes more, not less, important as a guardrail against document flooding and citation fabrication.

For Indian family law practitioners, the case provides a template for addressing a scenario that will increasingly arise: a self-represented litigant on the other side who has used AI to draft their pleadings. Indian courts may not yet have published AI hallucination rulings, but the foundational principles — truthfulness, candour, and accuracy of legal citations — are deeply embedded in the BCI Rules and CPC. Practitioners who identify fabricated citations in opposing documents should promptly raise the issue through a formal objection or an application under Order XVII Rule 3 CPC, building an evidentiary record that could support a costs order or striking of the defective pleading.