More than 50 of roughly 2,250 published federal appellate opinions issued between January and early August 2026 show signs of AI-generated prose, according to an analysis by Josh Morrow, a partner at Lehotsky Cohn LLP and a Harvard Law School graduate. Morrow used the AI-detection tool Pangram to analyze all published circuit court opinions from that period, with results ranging from under 1% to more than 50% AI-written text within individual opinions.
Most opinions clustered near the lower end of that range, and Morrow does not allege any specific judge or opinion is deficient as a matter of law. But his findings give the first systematic picture of how AI text and judicial drafting standards mix, and they raise questions the profession must address.
What the Data Shows
To establish a baseline, Morrow tested all published circuit opinions from January 2022, a period predating widespread large language model use. Pangram returned 0.000000% AI-generated text for every single one of those opinions. The contrast with 2026 was stark.
Pangram claims a false positive rate of 0.0041%, or roughly one false positive for every 24,000 documents. An independent evaluation by the University of Chicago Booth School of Business verified a false positive rate of 1 in 10,000 overall, dropping to 1 in 25,000 on an academic essay dataset comparable to Turnitin's own evaluation set.
In August 2025, Booth researchers Brian Jabarian and Alex Emi published a paper titled "Artificial Writing and Automated Detection." They concluded that Pangram is "the only detector that meets a stringent policy cap (False Positive Rates ≤ 0.005) without compromising the ability to accurately detect AI text." That standard matches the threshold applied under institutional AI-integrity policies.
The tool's latest version, Pangram 4, correctly identifies 99.66% of AI-generated documents and achieves a false negative rate of 0.3396%. That represents nearly six times fewer false negatives and fourteen times fewer false positives than its predecessor. Domain-specific data show false positive rates of 0.01% for creative writing, 0.02% for academic writing, and 0.01% for biomedical writing, with the tool performing less reliably on niche genres such as poetry.
Morrow uploaded the full text of each opinion to Pangram, then re-ran those with substantial AI signals from a separate account to guard against noise. The duplicate results were consistent, which he treats as evidence that Pangram's classifications were stable rather than random.
A Profession Already Moving Towards AI
Morrow's findings land in a context where AI adoption is already well under way. A Northwestern University survey of 502 randomly selected federal judges, conducted in December 2025 and published by the Sedona Conference in March 2026, found that over 60% of responding judges use at least one AI tool in their chambers, primarily for legal research and document review. Nearly half reported receiving no AI training from their court.
Over 300 federal judges across every circuit have issued standing orders or local rules addressing AI use in filings, according to a survey of current court AI rules. Those approaches generally fall into disclosure requirements or cautious-use guidelines. But those rules govern litigants' submissions, not what judges and law clerks may use when drafting opinions.
Morrow's position is that AI can improve legal writing and reasoning. He does not suggest that flagged opinions are deficient. The structural concern he raises is more specific: writing an opinion is part of the judicial function, not merely a record of a conclusion already reached. If extended passages arrive essentially finished from a language model and remain largely unedited, he argues, a question arises about whether the discipline that writing imposes on legal reasoning has been displaced.
He is also careful about the limits of his evidence. Pangram analyses only final text; AI used for research, outlining, or reviewing a draft leaves little trace in the published opinion. Human editing can erode the signals Pangram relies on. The tool cannot identify how AI-generated language entered the drafting process, nor reconstruct whatever judicial review followed.
The Unresolved Questions
The normative questions are ones the profession has not yet resolved: when, if ever, should AI involvement in judicial drafting be disclosed, and to what degree does apparent AI authorship affect an opinion's persuasive or precedential weight? At least two federal district judges have already publicly acknowledged AI-generated text in their opinions, and litigants have raised the issue in state proceedings.
Morrow's data does not answer those questions. But it does move them from theoretical to practical, and it suggests the appellate courts will need a disclosure framework and standard before detection tools make that conversation unavoidable.
Why this matters for legal professionals
For lawyers and litigators, the immediate takeaway is practical. If a circuit opinion shows signs of AI-generated prose, its analytical strength is not necessarily undermined. But when a court's reasoning process is partly opaque, the persuasive value of that opinion becomes harder to gauge - and precedent built on under-analyzed text is a weaker foundation for your next argument. A practical move: before relying on a recent unpublished or non-precedential decision, run it through a reliable detector with verified low false-positive rates, and read flagged sections for actual reasoning quality rather than rhetorical fluency.
Your membership also unlocks:
