A.K. v. M.R.: Indiana Court Admonishes Pro Se Appellant for AI-Generated Reporter Citation Mismatches

Background: A Neighbor Dispute and a Pro Se Appellate Brief

A.K., proceeding pro se, sought a civil protective order against his neighbor M.R., alleging stalking and repeated harassment. After an evidentiary hearing, the trial court denied the petition, finding the evidence insufficient. A.K. appealed, filing a brief that included multiple case citations to the North Eastern Reporter — Indiana’s regional reporter — but many of these citations did not match the case names given, and cited cases did not support the legal propositions A.K. attributed to them.

The Indiana Court of Appeals affirmed the trial court’s denial on the merits. But before reaching the substantive analysis, the court addressed the citation problems at length.

The AI Issue: Mismatched Reporter Citations and Unsupported Propositions

The court identified two specific problems with A.K.’s brief: (1) authorities cited did not support the propositions claimed, and (2) multiple North Eastern Reporter citations did not correspond to the case names in the brief. The court noted this pattern “could be chalked up to mere carelessness” but concluded it was “more likely” the result of legal research using generative AI — a probabilistic judicial inference now becoming standard practice.

The court cited its own recent precedent, Williams v. Kirch, 268 N.E.3d 284 (Ind. Ct. App. 2025), which had described citations to fictitious, AI-generated authority as “a growing problem nationwide” and noted that courts had sanctioned both attorneys and pro se litigants for including them. Because M.R. did not request sanctions, the court deemed an admonishment sufficient.

Holdings

1. Protective order denial affirmed. A.K.’s evidence was insufficient to establish stalking under Indiana law.
2. AI admonishment issued. A.K. was admonished for problematic citations and cautioned against using AI for legal research without independently verifying citations generated.
3. Indiana standard confirmed: Williams v. Kirch established that judges must rely on authentic citations — this principle applies equally to pro se appellants.

“We think it more likely that they are a result of legal research using generative artificial intelligence… Judges must be able to rely on the authenticity of the authorities cited by the parties to make just decisions.”

— Indiana Court of Appeals, A.K. v. M.R., March 10, 2026

India Angle: AI in Domestic Violence and Protection Order Proceedings

India’s Protection of Women from Domestic Violence Act 2005 and the CrPC provide mechanisms for protection orders that are frequently sought by self-represented complainants. As AI tools make brief drafting easier, AI-assisted applications in domestic proceedings will increase — with the risk that fabricated citations in such applications could undermine otherwise legitimate protection claims.

Relevant Indian Law

• Protection of Women from Domestic Violence Act 2005: Applications for protection orders, residence orders, and monetary relief are often filed by the complainant directly or through a Protection Officer. AI-generated applications containing false citations could be struck under Section 23 (ex parte orders) and could harm the credibility of genuine victims.
• BCI Rule 9: Advocates assisting complainants must not submit false citations, even in urgent domestic proceedings where speed is prioritized over accuracy.

Three Practical Tips

1. In urgent protection proceedings, verify first — file second. The urgency of a protection order does not justify unverified AI citations. Fabricated citations can give the opposite party grounds to challenge the order’s legal basis.
2. Citation reporter mismatches are a red flag. If an AI tool provides a case name with a reporter citation and the cited volume does not contain that case, it is a hallucination. Always check the volume-page combination independently.
3. The “probably AI” judicial inference is now routine. Mismatched citations will lead courts to infer AI use — and an AI admonishment in a published Indian appellate order would permanently mar the case record.

Quick Takeaways

• Reporter citation mismatches (case name does not match the N.E.3d citation) are a distinctive AI hallucination pattern identifiable without full-text database search.
• Courts now draw an inference of AI use from citation patterns — no admission required.
• The Indiana Court of Appeals has an established AI citation standard via Williams v. Kirch that it actively applies.
• Where the opposing party does not request sanctions, admonishment is the proportionate response — but the admonishment is published and permanent.

Deep Dive: The Reporter Citation Mismatch as AI Fingerprint

The specific error pattern in A.K. v. M.R. — North Eastern Reporter citations that do not correspond to stated case names — is a particular AI hallucination signature. AI language models generate citation formats that look syntactically correct (e.g., “268 N.E.3d 284”) but combine case names and reporter citations that belong to different cases. This is different from total fabrication: the case name may be real, the reporter citation may be real, but they belong to different cases.

This pattern arises because AI models learn citation formats from training data that includes millions of legal citations. The model learns the pattern (Case Name, Volume Reporter Page) but does not always match the correct triplet. For Indian practitioners, the equivalent risk is citing a judgment with a correct SCC citation that belongs to a different case, or vice versa. A quick check of the SCC Online landing page for the cited volume and page would catch this immediately.