AI-Drafted Lawsuits Are Swamping Federal Courts, MIT Study Finds

A new study from MIT and the University of Southern California documents what every federal court clerk has been quietly saying for two years: AI tools like ChatGPT and Claude are fueling a massive surge in lawsuits filed by people representing themselves. Pro se filings — cases without an attorney — have nearly doubled since ChatGPT launched, jumping to 16.8 percent of all federal civil cases in fiscal year 2025 from a steady 11 percent for the previous two decades.

The numbers tell two stories at once. Access to the legal system, gatekept for decades by the cost of lawyers, is being democratized. And the courts, never designed to handle that volume of filings, are starting to buckle.

The Context: Why Pro Se Cases Used to Be Rare

For most of US history, filing a federal lawsuit without a lawyer was technically possible but practically rare. Federal procedure is complex. The pleading requirements are exacting. Judges routinely dismissed pro se cases for failure to state a claim under the technical rules of Federal Rule of Civil Procedure 12(b)(6), regardless of whether the underlying grievance had merit.

The result was a two-tier justice system. People who could afford counsel had access; people who couldn't, didn't. The pro se rate hovered at 11 percent for two decades, mostly composed of prisoners filing constitutional claims and individuals so motivated they were willing to learn federal procedure on their own.

ChatGPT launched in November 2022. By fiscal year 2025, the pro se rate had jumped to 16.8 percent — 41,490 filings in a single year, nearly double the pre-AI average. The MIT-USC researchers ran the obvious test: they used text-detection tools to estimate how much of the language in pro se complaints was AI-generated.

The Details: AI Text Goes From 1% to 18% in Three Years

The trajectory the researchers documented is steep. The share of pro se complaints containing AI-generated text climbed from 1.0 percent in 2023 to 3.5 percent in 2024 to 10.5 percent in 2025 to 18.0 percent in early 2026. That's not a wave — it's a structural shift in how a meaningful slice of Americans now access the legal system.

The case quality story is also more nuanced than the early skepticism suggested. Many AI-drafted complaints are technically competent: they cite the right statutes, identify the right legal theories, and survive initial procedural screening that previously killed similar pro se attempts. Real people are winning real cases.

Amr Abouelmagd, for example, took a cross-border child-custody case from federal court in Brooklyn all the way to the Supreme Court without a lawyer, relying on a mix of AI advice, court clerks, and friends. The case is one of dozens that practitioners now point to when arguing that AI is meaningfully expanding access to justice.

But case duration tells a different story. Docket activity in pro se cases rose 158 percent, even though average case duration didn't increase proportionally. That means courts are processing far more filings per case to reach the same outcome. The procedural complexity that used to filter out weak cases at the pleading stage is now playing out across the full life of the case.

Industry Impact: Courts, Lawyers, and Legal Tech

For federal courts, the math is brutal. Some district courts are reporting 20 to 40 percent increases in pro se civil filings. Court budgets, judge counts, and clerk staffing are set by Congress through appropriations that move slowly. The system was designed for the old volume.

The Administrative Office of the US Courts has been quietly raising the alarm for over a year. MIT researcher Anand Shah, one of the study's authors, warned that AI-driven filings are creating workloads "that could swamp courts" if filing growth continues at the current pace. Several federal districts have explored new screening procedures specifically for AI-generated complaints.

For the legal profession, the implications cut multiple ways. The bottom of the market — small-claims work, simple landlord-tenant disputes, basic family law matters — is being eroded by AI-assisted self-representation. Junior associate work at firms that handled volume-based practices is shrinking. But complex litigation, where strategic judgment matters more than document production, is largely untouched.

Why this matters: the legal industry has been the highest-margin professional services sector for decades, partly because the procedural complexity required expensive expertise to navigate. AI just collapsed part of that moat. The work that survives is the work that requires actual legal reasoning, not just document drafting.

Expert Perspectives

Access-to-justice advocates see this as the clearest example yet of AI delivering on its promise. The American Bar Association's longstanding "justice gap" — the estimated 92 percent of low-income civil legal needs that go unmet — has been a structural failure of the legal system for generations. If AI is helping a measurable share of those people draft viable filings, that's a major social good.

Skeptics emphasize the quality problem. Some AI-generated complaints contain "hallucinated" case citations — citations to court decisions that don't exist. Several pro se litigants have been sanctioned for filing briefs containing fictional precedent. The same accessibility that helps legitimate plaintiffs also enables frivolous and vexatious filings at scale.

Judges, predictably, are split. Some have publicly praised the improved quality of pro se filings since ChatGPT. Others have expressed concern that AI is producing "form" complaints that mimic the structure of legitimate cases without the substance, requiring the court to do more work to identify what's real.

What's Next

Expect two responses to develop over the next 18 months. First, court rule changes. Federal districts will likely adopt local rules requiring disclosure of AI-assistance in pro se filings, similar to disclosure rules already in place for represented parties. Several districts are also piloting AI-assisted triage tools to handle the increased volume.

Second, a shift in legal-tech investment. Startups that built tools assuming lawyers were the customer are now competing with tools aimed directly at self-represented litigants. State courts handle a much larger civil volume than federal courts, and several state systems are seeing similar pro se surges — California, New York, and Texas are likely to move first on rules and tooling.

Bottom Line

AI is doing exactly what its proponents promised — and exactly what its skeptics feared — to one of America's most exclusive professional gatekeepers. People who couldn't afford a lawyer can now file viable lawsuits. Courts that weren't designed for that volume are straining. The next two years will determine whether the legal system adapts or whether the pro se surge forces a deeper rethink of how civil justice actually works. If you've ever wanted to take someone to court but couldn't afford a lawyer, you should know: the math just changed.