AI Hallucinations in Court Filings Drop; Subtle Errors Rise in 2026

3 min readSources: LegalTech News

AI hallucinations in court filings fell for two quarters in 2026, while subtle errors grew.

Why it matters: Legal professionals must adapt as AI errors evolve from obvious fake citations to harder-to-detect inaccuracies, maintaining diligence to avoid misleading courts and sanctions.

  • AI hallucinations in court filings dropped for the second consecutive quarter in 2026.
  • Researcher Damien Charlotin tracked more than 1,500 AI-related errors in U.S. courts by October 2026.
  • Nearly 900 court rulings cited filings flagged for AI errors, including 603 in U.S. courts as of November 2026.
  • A 2025 study found 146,932 AI-hallucinated citations in 2.5 million scientific papers, highlighting widescale fabrication risks.

AI-generated fake citations in court filings have declined notably in 2026, marking an important development in legal technology. According to Law.com, filings with AI hallucinations dropped for two straight quarters, suggesting initial improvements in AI output quality or increased scrutiny by legal professionals.

Yet, while overt fake references have decreased, subtler AI errors—such as inaccurate paraphrasing or partial misrepresentations of case law—are on the rise. Damien Charlotin, a legal researcher tracking AI errors in judicial documents, documented over 1,500 such instances in U.S. state and federal courts by October 2026. His continuous monitoring highlights that these nuanced inaccuracies often evade quick detection, increasing the risk of misleading judges and attorneys. Direct access to Charlotin’s reports remains limited, but his data has been cited in multiple legal technology analyses.

The ongoing impact is evident in court decisions: nearly 900 rulings as of November 2026 reference filings flagged for AI hallucinations, including 603 in U.S. courts alone. This suggests courts remain highly attentive to AI-generated errors, which can affect case outcomes.

Parallel concerns arise beyond legal filings. A 2025 study analyzing 111 million scientific references across 2.5 million papers found 146,932 AI-hallucinated citations. This study confirms the broader risk of AI generating fabricated data or sources, beyond the courtroom.

Legal regulators have responded. In October 2025, the Illinois Attorney Registration and Disciplinary Commission released The Illinois Attorney’s Guide to Implementing AI, advising lawyers on responsible AI use to avoid ethical violations stemming from hallucinations.

Judicial voices emphasize the issue. U.S. District Judge John Tuchi publicly highlighted prevalence of fabricated AI citations in filings he reviewed, confirming AI hallucinations’ significant courtroom footprint. Legal commentator Zach Warren noted sanctions remain common for filings containing bogus AI-generated case citations, reinforcing accountability.

These developments underscore the need for legal professionals to rigorously verify AI-generated content. Maintaining accuracy ensures integrity in filings and mitigates risks of judicial sanction or reputational harm.

By the numbers:

  • 2 — consecutive quarters in 2026 with declining AI hallucinations in court filings
  • 1,500+ — AI-related errors tracked by Damien Charlotin in U.S. courts as of October 2026
  • 603 — U.S. court rulings citing filings flagged for AI errors by November 2026
  • 146,932 — hallucinated citations found in scientific papers in a 2025 study

Yes, but: While overt fake citations are decreasing, the rise in subtle errors could prove more challenging to detect and address, requiring enhanced vigilance.

What's next: Legal regulators and courts are expected to update guidelines and sanctions as AI use becomes more integrated into legal drafting and research processes.