AI is supercharging self-filed lawsuits, and courts are scrambling

AI is not just helping people write emails and pitch decks. It is now helping them sue, and the federal courts are feeling it. A new study examining 4.5 million federal civil cases from 2005 to 2026 found that the share of lawsuits brought by self-represented people jumped from 11% in 2022 to 16.8% in 2025, and within those cases the number of filings made more than doubled compared with before 2023. Judges are seeing the change in real time, and many of them think AI is a big reason why.

Judge Maritza Braswell, a federal magistrate judge in Colorado, is one of them. She spends her days reading stacks of documents written by people without lawyers, and she says she can spot large language model writing by the prose, and sometimes by hallucinated cases and fabricated quotes. “I do correlate that to AI in part because I see AI use,” she says. But this is not a simple story of bots flooding the courthouse with junk. Braswell says she is also seeing better-drafted pleadings. And while she notes that some filings contain errors, she can often understand the arguments more clearly when AI has helped draft them than when the litigant wrote them alone. In other words, AI is not only increasing volume, it is changing the quality of what judges have to process.

The study tried to test that hunch with data, not vibes. Authors Anand Shah at MIT and Joshua Levy at USC ran 1,600 randomly sampled court documents through Pangram, a commercial AI-text detector. The share flagged as containing AI-generated writing rose from 1% in 2023 to 18% in 2026. That lines up with what judges like Braswell are seeing from the bench. It also explains why some courts may actually welcome the change, at least in part. Historically, filings from people without lawyers have often been hard to decipher, including handwritten scrawls bordering on gibberish. Judges had to read them charitably and spend extra time figuring out what the litigant was even asking for. AI-drafted motions, by contrast, are often cleaner, faster to parse, and easier to rule on. That matters because court systems are already strained, and anything that reduces decoding time can affect throughput.

Still, more readable does not mean more winnable. The study found that people without lawyers are still far more likely to lose than people with lawyers, and AI has not changed that. Joshua Levy put the point bluntly: “It turns out that mounting a lawsuit is a complex, multifaceted task. Not all of it is just drafting text.” That distinction matters for executives because it is the gap between automation that improves presentation and automation that changes outcomes. AI can make a claim look more professional without making it stronger. It can also help nonlawyers feel more confident than they should, which may raise the number of filings without improving the underlying odds.

That confidence is already spreading online. Communities are trading self-help guides on how to use AI to sue. In December 2024, a viral Reddit post walked immigration applicants through suing the United States Citizenship and Immigration Services over delayed review of their applications: draft a writ of mandamus with Microsoft Copilot, pay a lawyer $150 to polish it, and file in the expedient District of Vermont. The impact was immediate. Cases filed by people without lawyers in Vermont rose from about 45 a year before 2022 to more than 1,100 in 2024. That is a huge jump, and it shows how quickly a simple workflow, a chatbot, and a social media post can alter behavior at scale. For leaders, it is a reminder that distribution can matter as much as technology. The tool is one thing. The viral playbook is what turns it into a system.

The legal system is now wrestling with a deeper question: what, exactly, is the relationship between a person and the chatbot helping them act like a lawyer? Judge William Garfinkel, a federal magistrate judge in Connecticut, has been on the bench for three decades and is now asking whether conversations with chatbots dispensing legal advice should be privileged the way lawyer-client conversations are. “You can make a good argument that... conversations with large language models like Claude or ChatGPT or Grok should deserve some protection,” he says. Courts are already split. In February, a federal court in Michigan ruled that a self-represented person’s conversations with ChatGPT to prepare her case were work product, meaning legal work shielded from the other side. The same day, a federal court in New York held that documents a criminal defendant generated using Claude were not privileged attorney-client conversations or work product. That court said Claude is not an attorney and that the user had no “reasonable expectation of confidentiality in his communication” with Claude because AI companies could disclose user data to third parties. In March, Braswell also ruled that a self-represented person’s use of a chatbot should stay off limits, writing: “It is true that AI systems like ChatGPT, Claude, Gemini, and others... collect user data for training and other purposes. But... that does not eliminate all expectations of privacy.” Courts have since remained split on the issue.

The next fight may be liability. Judge Allison Goddard, a federal magistrate judge in California, says she has noticed people without lawyers getting bad settlement advice from ChatGPT, including one plaintiff who slipped and fell in a store and asked for $700,000, a number the judge said was wildly higher than the case was worth. “Where are you getting the idea that you’re getting $700,000? Did you go to ChatGPT?” she asked. “Well...” the plaintiff mumbled. Goddard then walked the plaintiff through the law and suggested a lower amount. “It’s like Dr. Google went to law school,” she says. That may be funny in the room, but it is becoming a board-level question outside it. In March, Nippon Life Insurance Company sued OpenAI alleging that ChatGPT practiced law without license and helped a woman reopen a lawsuit that was already settled, flooding the court with frivolous filings. “ChatGPT is not an attorney,” the lawsuit said. In May, OpenAI asked the court to dismiss the case, arguing that ChatGPT does not practice law. “ChatGPT is not a person and neither has nor uses any degree of legal knowledge or skill,” OpenAI said in its filing. The case is still pending before the court.

Lawmakers are starting to move too. New York introduced a bill in March that would bar chatbots from impersonating lawyers, even if they notify users that they are interacting with chatbots. In Congress, a series of bills have been proposed to ban chatbots from posing as lawyers, doctors, and other licensed professionals, but they have yet to gain traction. For now, though, people will keep using AI as their lawyer, especially when the system feels too intimidating to face alone. Braswell says that with AI, a self-represented litigant now answers her questions confidently after rehearsing with a chatbot. “This is a really tough system to navigate. With AI, though, it gets a little less complex,” she says. That is the real story here: AI is lowering the barrier to entry, even as the courts are still figuring out who bears the risk when the advice is wrong.