How courts are coping with a flood of AI-generated lawsuits

Federal magistrate judge Maritza Braswell in Colorado has witnessed a dramatic transformation in her daily workflow over the past eighteen months. Each morning, she encounters an expanding volume of court filings submitted by self-represented litigants, many of whom lack legal representation due to financial constraints or the marginal nature of their claims. A comprehensive study examining 4.5 million federal civil cases spanning from 2005 through 2026 documents this shift with precision: self-represented filings climbed from eleven percent of all cases in 2022 to 16.8 percent by 2025, with the absolute number of such submissions more than doubling compared to the pre-2023 baseline. Judge Braswell attributes this surge directly to the proliferation of artificial intelligence tools available to ordinary citizens, a conclusion she reaches not through speculation but through her direct observation of distinctive writing patterns and, troublingly, the recurring appearance of fabricated case citations and invented quotations that betray AI authorship.

The emergence of AI-assisted legal filing represents a watershed moment in access to justice infrastructure. Prior to the recent explosion of generative AI capabilities, individuals unable to afford lawyers faced formidable barriers when navigating complex federal procedures and documentary requirements. The traditional route of pro bono legal assistance has remained severely constrained by resource limitations, leaving millions of Americans effectively locked out of the court system. The democratization of writing tools through large language models has fundamentally altered this calculus, providing instantaneous assistance with document drafting that previously required either professional legal training or substantial financial investment. Yet this development arrives amid deeper questions about judicial capacity, attorney professional responsibility, and whether expanded filings without corresponding improvements in actual case outcomes constitute genuine progress or merely the distribution of failure across a wider population.

The empirical evidence supporting AI's role in this transformation comes from rigorous testing conducted by researchers Anand Shah at MIT and Joshua Levy at USC. They subjected 1,600 randomly sampled court documents to Pangram, a commercial artificial intelligence text detection system, which revealed that the proportion of filings flagged as containing AI-generated content escalated from one percent in 2023 to eighteen percent by 2026. This eighteenfold increase over a mere three-year window establishes a clear temporal correlation with the public release and adoption of advanced language models like ChatGPT and similar systems. Simultaneously, the structural characteristics of self-represented filings have demonstrably improved in readability and organizational coherence compared to the handwritten submissions and incoherent documents that traditionally dominated this category. Judge Braswell's assessment that she can process AI-drafted motions more rapidly than those composed without algorithmic assistance suggests measurable efficiency gains within judicial operations, despite the need for heightened scrutiny regarding factual accuracy.

The practical implications for litigants and the judiciary extend beyond mere convenience or workflow efficiency. For individuals operating without legal counsel, the availability of AI tools represents the difference between filing incomprehensible documents that judges struggle to interpret and producing submissions that clearly articulate legal positions and procedural arguments. Judge Braswell's observation that clearer pleadings enable her to rule more effectively acknowledges a fundamental axiom of judicial review: a judge cannot fairly adjudicate claims they cannot comprehend. Yet this accessibility gain arrives with significant caveats. The enhanced clarity of AI-drafted documents masks a critical problem: these same tools demonstrably fabricate legal citations, invent case names, and create quotations from non-existent sources, introducing sophisticated falsehoods into the record that less experienced judges might overlook. The consequence is that while self-represented litigants may now file more intelligible submissions, those submissions carry embedded risks of deception that could mislead courts or opposing parties. Moreover, the data reveals a troubling reality that undercuts the narrative of improved access: despite writing better-drafted pleadings, self-represented litigants win no more often than they did previously, suggesting that superior documentation cannot overcome the inherent disadvantages of lacking substantive legal expertise and courtroom experience.

This phenomenon illuminates a broader pattern within the AI revolution: technology often redistributes problems rather than solving them fundamentally. The surge in AI-assisted filings concentrates new burdens on judicial systems already operating under resource constraints, while simultaneously creating novel questions about legal liability and responsibility. Courts must now determine whether chatbots function as tools for which users bear full responsibility, or whether language models themselves incur duties parallel to those governing human attorneys. Legislators across multiple states are beginning to address who bears financial liability when artificial intelligence systems provide demonstrably deficient legal guidance that damages users' cases or rights. This governance challenge reflects a deeper tension in deploying AI at scale: the technology expands capability and access without automatically improving outcomes or clarifying accountability structures. The pattern evident in federal courts mirrors similar dynamics across other sectors where AI has accelerated the volume of human interaction with institutions while complicating the assignment of responsibility for errors, deceptions, or systemic failures.

Stakeholders should monitor several specific developments that will shape how courts ultimately accommodate this shift. The American Bar Association and state bar associations across the country are actively developing ethical guidelines addressing AI-assisted legal work, with particular attention to the September 2025 timeframe when revised professional responsibility rules are expected to emerge from multiple state bodies. Federal court administrators, coordinating through the Administrative Office of the United States Courts, have initiated working groups to establish protocols for identifying and handling AI-generated documents with embedded hallucinations. Additionally, litigation specialists should track pending legislation in California, New York, and Illinois that would establish liability frameworks for AI systems providing legal guidance, with proposed provisions potentially impacting how courts treat AI-assisted pro se filings. Judge Braswell's courtroom will likely serve as a testing ground for practical approaches, and decisions she issues regarding the admissibility of AI-flagged content and the handling of fabricated citations may establish precedents influencing how federal judges nationwide approach this expanding category of submissions.