Court Clerks Are About to Become AI Fact-Checkers

Something peculiar is happening in federal courthouses across America. Self-represented litigants—people filing lawsuits without lawyers—have discovered ChatGPT, and they're using it to draft their legal complaints. In 2023, about 1% of court filings showed evidence of AI generation. By 2026, that number jumped to 18%. That's not a trend. That's a revolution in how ordinary people access the legal system.

What makes this particularly interesting is the reaction from judges themselves. According to recent reports, many judges find AI-drafted pleadings clearer and easier to understand than traditional handwritten documents from self-represented parties. Think about what that means: a technology that legal experts spent the past two years panicking about is apparently making their jobs easier when it comes to processing the most challenging category of filings they receive.

But here's where it gets complicated. The same week we learned about courts struggling with this flood of AI-generated lawsuits, Meta's AI customer support agent was exploited to steal Instagram accounts through simple social engineering. The AI blindly followed instructions to change email addresses without proper security verification—exactly the kind of mindless compliance that makes people nervous about AI in high-stakes environments like courtrooms.

This tension captures something fundamental about where we are with AI deployment in 2026. The technology is simultaneously too good and not good enough. It's good enough that non-lawyers can use it to produce coherent legal documents that judges prefer reading. It's not good enough to recognize when it's being manipulated or to exercise the judgment that distinguishes a legitimate request from social engineering.

The legal system is inadvertently becoming a testing ground for a question that applies far beyond courtrooms: what happens when AI makes a gatekept professional service suddenly accessible to everyone? We've seen hints of this in other domains—AI-designed vaccines entering human trials, chatbots handling customer support—but the legal system offers unique clarity because court records are public and judges have to publicly explain their decisions.

Right now, courts are adapting by treating AI-generated filings essentially the same as human-generated ones, judging them on substance rather than origin. That's a pragmatic response, but it kicks a harder question down the road: if AI can produce legal filings that judges prefer to human-written ones, what does that say about the barriers we've built around legal expertise? And if those barriers fall, who ensures the AI isn't just producing clearer nonsense?

The early evidence suggests judges are developing new skills—essentially becoming AI fact-checkers rather than just legal arbiters. They're learning to spot the telltale signs of chatbot generation and to probe more carefully when filings seem suspiciously polished compared to a litigant's apparent sophistication level. Court clerks are adapting too, developing informal protocols for handling the new flood of filings.

What we're witnessing isn't courts coping with AI. It's courts being quietly transformed by AI into something different—more accessible, perhaps more efficient, but requiring new competencies from everyone involved. The judges who once spent hours deciphering handwritten complaints are now spending that time verifying that coherent-sounding AI-generated arguments actually make legal sense.

Nobody designed this experiment. It's just happening, one ChatGPT-drafted complaint at a time. And unlike the controlled rollouts we see in robotics or medicine, there's no pause button. The legal system will adapt because it has to. The question is whether other institutions are watching and learning, or whether they'll be just as surprised when AI comes for their gatekeeping functions next.