Generative AI is beginning to reshape the U.S. legal system by dramatically lowering the barrier to filing lawsuits without a lawyer, according to a new working paper from researchers at Massachusetts Institute of Technology and University of Southern California.
The paper, Access to Justice in the Age of AI: Evidence from U.S. Federal Courts, analyzed more than 4.5 million federal civil cases and 46 million PACER court records spanning 2005 through early 2026.
Researchers found that the share of federal civil lawsuits filed by self-represented litigants, known as pro se cases, jumped from a long-running baseline of roughly 11% to 16.8% in 2025.
The increase marks the first major structural break in pro se litigation rates in more than two decades.
AI Lowers the Cost of Filing Lawsuits
The study argues that modern large language models are making it dramatically easier for ordinary people to navigate legal procedures, draft complaints, and initiate lawsuits.
According to the authors, AI systems particularly help in “formulaic” case types where success depends heavily on producing correctly structured legal documents.
The strongest growth appeared in areas such as:
- Civil rights complaints
- Consumer credit disputes
- Foreclosure proceedings
- Employment-related claims
More specialized and attorney-intensive litigation — including patent disputes and securities fraud — showed little change.
The researchers argue that generative AI effectively reduces the “cost of entry” into the legal system by automating work that previously required legal expertise.
Court Activity Is Rising Sharply
The study found that the rise in AI-assisted litigation is already increasing operational pressure on federal courts.
While self-represented cases are not resolving faster than before, the amount of activity within each case has surged.
By 2025, the number of docket entries generated by pro se litigants during the first 180 days of a case rose 158% above pre-AI levels.
That includes:
- Motions
- Responses
- Procedural filings
- Judicial orders
- Scheduling actions
The researchers argue this matters because every filing consumes judge and court staff time in a system where adjudication capacity is already constrained.
“Each docket entry is a claim on the court’s time,” the paper states.
Researchers Found AI-Generated Text in Court Filings
To directly test whether AI tools were appearing inside legal documents, the authors sampled 1,600 federal civil complaints filed between 2019 and 2026.
Using AI-text detection tools, they found virtually no AI-generated filings before widespread LLM adoption.
That changed rapidly after the release of modern generative AI systems.
According to the paper:
- Around 1% of complaints contained detectable AI-generated text in 2023
- The share rose to 10.5% in 2025
- By early 2026, roughly 18% of sampled complaints showed signs of AI-generated writing
The authors caution that the real number could be higher because publicly available court-text archives likely underrepresent self-filed cases.
Federal Courts May Be an Early Warning Sign
The paper argues the federal system may only represent the beginning of a broader transformation.
Federal courts are among the hardest venues for self-represented litigants because they require complex jurisdictional and procedural compliance.
If AI is enabling successful self-representation there, researchers believe the impact in state and municipal courts – where most U.S. litigation occurs – could be substantially larger.
The authors also warn that courts currently lack scalable mechanisms to absorb rapidly rising litigation volume.
Unlike other sectors disrupted by AI, the supply of federal judges cannot quickly expand, and federal courts currently prohibit judges from using AI systems to draft opinions.
AI Is Reshaping Institutional Systems, Not Just Workflows
The paper frames the legal system as one of the first major public institutions experiencing measurable downstream effects from widespread generative AI adoption.
Rather than simply increasing worker productivity, the researchers argue AI is beginning to alter who participates in institutional systems in the first place.
The study concludes that courts now face a broader structural question: whether AI becomes a productivity tool that improves access to justice, or a source of systemic congestion that overwhelms institutions built around human review and judgment.
