
More lawsuits in California are being written with help from AI chatbots, especially by people who go to court without a lawyer, according to a study released on November 19 by researchers at Stanford University. The team, which includes law professor Daniel E. Ho, examined 127,518 complaints, the documents that open a civil lawsuit, filed in San Francisco and Santa Clara counties in California and in Oklahoma and Tulsa counties in Oklahoma. Leaving out debt collection cases, which follow fixed templates, they estimate that 38% of complaints in Santa Clara and 36% in San Francisco show signs of AI drafting in 2026, against 16 to 17% in the Oklahoma counties. Before 2023 the share was close to zero. Among people representing themselves in Santa Clara, it rose from 19% in 2024 to 72%; among complaints filed by attorneys, it is about a third.
Measuring this is difficult. Standard AI detectors stumble on legal writing, which is full of stock phrases, and their results swing widely depending on their settings. So the researchers gathered 980 filings in which judges had caught invented legal citations, a sure sign of AI use, and used them to calibrate their estimates. They warn that the method shows trends across thousands of cases but cannot prove whether any single filing used AI. The authors ask courts for clear disclosure rules. A new California law, SB 574, will require lawyers to disclose AI use, but according to KQED it does not cover people who file on their own.