Thursday, November 19, 2026·Focal News

The San Francisco Focal

San Francisco's independent voice

Local

Stanford Study Finds AI Use in Bay Area Civil Lawsuits Surged to Over a Third This Year

New Stanford research shows more than a third of non-debt civil lawsuits filed in Santa Clara and San Francisco counties this year bear signs of AI-assisted drafting, with self-represented litigants driving the steepest increase. The findings raise questions about access to justice, court capacity, and the risk of fabricated citations.

Stanford Study Finds AI Use in Bay Area Civil Lawsuits Surged to Over a Third This Year

More than one-third of non-debt civil lawsuits filed this year in Santa Clara and San Francisco counties show signs of AI-assisted writing, according to a new Stanford University study that offers the first systematic look at how generative AI is entering state trial courts.

The report, from Stanford's Human-Centered Artificial Intelligence institute and its Regulation, Evaluation, and Governance lab, found the sharpest rise among people representing themselves. In Santa Clara County, the share of complaints filed by self-represented litigants showing signs of AI jumped from 19% in 2024 to 72% this year. About one-third of attorney-filed complaints in the county also showed signs of AI use.

"The sharp rise we document here corroborates that AI in the law is a question of how, not whether," the researchers wrote.

The study is the first to measure AI's presence in state trial courts, which handle nearly all cases in the country. Researchers had to overcome a major obstacle: existing AI detectors trained on general language struggle with legal documents' specialized vocabulary and boilerplate language. To calibrate their tools, they used nearly 1,000 real court filings in which judges had flagged fake citations.

The findings come as courts across California and beyond grapple with legal citations hallucinated by artificial intelligence. The study's authors warn that detection will only become harder "as language models are trained to mimic human text and style."

Dan Ho, a Stanford law professor and co-author of the study, said there is no putting the horse back in the barn, even if some judges would prefer that neither civilians nor lawyers use AI. He worries about subtler damage to the law beyond fabricated citations.

"It isn't just hallucinations that you worry about," Ho said. "You may worry about the way in which generative AI might not hallucinate directly a case, but still craft a kind of argument that really shouldn't be relied upon."

Ho noted that many self-represented litigants are using generative AI to replace internet searches or library visits, and that the technology "could improve the quality of complaints, relative to not having much access to legal knowledge at all."

Riana Pfefferkorn, a Stanford HAI policy fellow who was not involved in the study, cautioned that "the mere use of AI in drafting a document doesn't imply anything about quality." She said the study points to a potential area for future research.

"If AI helps litigants draft legal documents but doesn't help them win, then litigants will really need to adjust their expectations about the capabilities and limitations of AI tools," Pfefferkorn said.

The study authors framed the development as a double-edged sword for a state already facing an access-to-justice crisis. "AI assistance could lower barriers to legal expertise and access to the justice system, but it could also strain a court system that is already at capacity by increasing the volume of cases, the length of documents and likelihood that citations and facts are fictitious," they wrote.

California courts already have guidelines for how judges themselves use AI, and a new state law will require lawyers to disclose AI use starting in January, though it does not apply to people representing themselves.

The report's authors suggested courts may eventually have to respond forcefully to errors by lawyers who fail to verify AI-generated content. "At the end of the day," one researcher wrote, "I suspect courts will have to come down hard on a string of errors by lawyers who fail to check some nonsense they get from AI."

More from The San Francisco Focal