In courthouses across the country, the sound of justice has always been paperwork: complaints, motions, exhibits, and the endless choreography of deadlines. For decades, courts have absorbed a steady stream of filings from self-represented litigants—people navigating legal systems without lawyers, often under stress and with limited resources. But a new variable is changing the rhythm. Increasingly, some plaintiffs are using artificial intelligence to draft lawsuits faster than they ever could before, producing filings that arrive in greater volume and, in some cases, with the kind of internal inconsistencies that clerks and judges must sort out before a case can move forward.
The result is not simply more work. It is a different kind of work—one that tests how courts triage, verify, and manage cases when the barrier to filing has dropped dramatically. When a person can generate a complaint in minutes, the question becomes less “Who has a claim?” and more “How do we process claims at scale without letting quality collapse?”
According to recent reporting, the shift is already visible in court dockets and in the day-to-day strain on court staff. The filings are sometimes described as “home-brewed,” a phrase that captures both the DIY nature of self-representation and the new reality of AI-assisted drafting. The concern is not that AI automatically produces fraudulent documents. It’s that AI can produce plausible-sounding legal narratives and procedural language that may be incomplete, inaccurate, or poorly grounded in the facts required for a viable case. And when those documents are generated quickly, errors multiply—sometimes across dozens of filings, sometimes across multiple jurisdictions, and sometimes in ways that are difficult to detect until a judge or clerk reads closely.
What makes this trend particularly challenging is that it collides with two competing goals that courts must balance. On one hand, access to justice matters. Courts are constitutionally and practically committed to allowing people to bring claims even when they cannot afford counsel. On the other hand, courts also have an obligation to manage their limited resources and to prevent the system from being overwhelmed by filings that do not meet basic standards of clarity, relevance, or legal sufficiency.
AI changes the math. It reduces the time and cost of drafting. It can also reduce the friction that typically discourages frivolous or poorly prepared filings—because the “blank page” problem is gone. A litigant no longer needs to know how to structure a complaint, how to cite statutes, or how to translate a grievance into legal terms. They can ask a model for a template, feed it details, and receive something that looks like a lawsuit. That appearance of legal formality can be powerful. It can also be misleading.
Court staff have long dealt with filings that are messy, repetitive, or difficult to interpret. But the new wave is different in its speed and volume. Clerks and judges are not just reading more; they are reading more that is generated with a tool designed to produce fluent text. Fluency can mask gaps. A complaint can read like it belongs in court while still failing to connect the facts to the elements of a claim. It can include citations that are technically formatted but not actually relevant. It can describe events in a way that conflicts with attached documents. It can also omit key information because the model assumes that what is missing is implied rather than required.
This is where the “bandwidth” metaphor becomes more than a headline. Court bandwidth is not only about storage or scanning. It is about attention—human attention. Every filing requires some level of review: screening for jurisdiction, assessing whether the complaint states a claim, checking whether required forms are included, determining whether service is properly requested, and deciding whether the case should proceed, be dismissed, or be corrected through orders. When filings increase, the bottleneck shifts from physical capacity to cognitive capacity. Clerks and judges must spend time untangling what should have been straightforward.
And because AI can generate variations quickly, the system can be flooded with near-duplicates. A litigant might file one complaint, get a dismissal or an order to amend, then return with a revised version that changes a few phrases while leaving the underlying defects intact. In traditional self-representation, revisions happen too—but usually at a slower pace, constrained by the litigant’s time, literacy, and ability to research. AI compresses that timeline. It can also encourage iterative filing strategies that test the limits of court patience: file, wait, revise, refile—until the system either absorbs the case or forces the litigant to stop.
That dynamic raises a deeper question: what does it mean to “manage” a docket when the volume is driven by a technology that can scale output? Courts have rules for frivolous litigation and sanctions for abuse, but those tools require identification and, often, a record. If the filings are numerous and inconsistent, building that record takes time. Meanwhile, the docket continues to fill.
There is also the issue of verification. In many legal contexts, accuracy is not optional. A complaint must identify parties correctly, allege facts that are within the plaintiff’s knowledge, and connect those facts to legal theories. AI-generated drafting can blur the line between what a litigant knows and what a model suggests. Even when a litigant provides the facts, the model may reorganize them, infer missing details, or smooth contradictions into a coherent narrative. That coherence can be persuasive, but it can also be wrong.
Consider a common scenario: a plaintiff attaches emails, screenshots, or a contract. The complaint may summarize those materials. If the summary is slightly off—if a date is wrong, if a quote is paraphrased inaccurately, if a clause is mischaracterized—the legal consequences can be significant. A claim might hinge on whether a notice was given, whether a deadline was met, or whether a particular representation was made. Small factual errors can undermine an entire theory. When AI is used, those errors can be introduced unintentionally and then repeated across filings.
The “home-brewed” aspect also points to a broader concern: the absence of professional review. Lawyers do not merely write. They check. They verify citations, confirm procedural requirements, and ensure that allegations are consistent with evidence. AI can draft, but it does not inherently verify. It can mimic legal reasoning, but it does not guarantee that the reasoning is anchored in the actual record. Without a human attorney to validate the work, the risk of error increases.
Yet it would be a mistake to frame the story as simply “AI is bad.” The more interesting—and more urgent—question is how courts should respond in a way that preserves access to justice while protecting the system from overload and from unreliable filings.
One approach is procedural triage: requiring clearer initial submissions, tightening screening standards, and using early case management to identify deficiencies quickly. Courts already do this to some extent. But AI-driven filings may require more targeted measures. For example, courts could require plaintiffs to provide specific factual statements tied to particular documents, rather than allowing broad narrative summaries. They could also require more explicit identification of what the plaintiff believes is true, what is based on records, and what is based on inference. That would not eliminate AI drafting, but it would force a separation between “text that sounds legal” and “facts that can be verified.”
Another approach is to adjust the burden of explanation. If a complaint is generated with AI, it may contain legal jargon that obscures rather than clarifies. Courts could respond by demanding plain-language factual allegations that answer concrete questions: who did what, when, where, and how it relates to the legal elements. This is not a new idea—courts have long asked for clarity—but AI may make clarity harder to achieve because the writing can be overly polished. Requiring structured factual answers could counteract that.
A third approach involves technology itself. If AI is increasing the volume of filings, courts may need tools to detect patterns—such as repeated templates, inconsistent citations, or anomalies in formatting and content. Some courts already use software to manage documents and track filings. The next step would be more sophisticated analytics that flag potential issues for human review. That raises its own concerns about fairness and transparency, but it also reflects a reality: if the system is being stressed by machine-generated text, the response may need to include machine assistance.
Still, any technological solution must be careful. Courts cannot treat all AI-assisted filings as suspect. Many legitimate pro se litigants may use AI to help with grammar, organization, or translation. The problem is not the existence of AI assistance; it is the combination of speed, scale, and insufficient verification. A complaint can be AI-assisted and still be accurate. The challenge is distinguishing between assistance and fabrication, between drafting support and substantive invention.
The reporting also suggests that the issue is not confined to a single type of case. AI can be used to draft everything from consumer disputes to employment claims to civil rights complaints. That breadth matters because it means the phenomenon is not limited to one niche area where courts might develop specialized procedures. Instead, it touches the core infrastructure of civil litigation.
There is also a political and cultural dimension. Self-representation has long been framed as a sign of access to justice problems: people cannot afford lawyers, so they file on their own. AI could be seen as a partial remedy—an inexpensive way to generate legal documents. But if AI also increases the number of filings that courts must process, it could paradoxically worsen the very access problem it aims to solve. A system overwhelmed by filings can become slower, more punitive toward procedural mistakes, and more likely to dismiss cases early. That would harm legitimate claims.
This is why the “unique take” on the story is not just about the volume of AI-generated lawsuits. It is about the feedback loop between technology and institutions. When AI lowers the cost of filing, more people file. When more people file, courts allocate more resources to screening and triage. When courts allocate more resources to screening, fewer resources remain for substantive adjudication. When substantive adjudication slows, litigants experience delays and frustration. Some may respond by filing more, including amended or