Generative AI and the Pro Se Litigant

By Alan S. Golub, Esq. and Robert E. Smithson, Jr., Esq.

The rise of generative AI has sent shockwaves through the legal profession. Hardly a week passes without a story of a major law firm filing briefs generated with the assistance of AI that cite nonexistent case law hallucinated by whatever large language model the firm happened to be using. Sanctions motions and disciplinary referrals over fabricated citations have become common enough that they barely make news outside the legal press anymore. Our profession has yet to fully come to grips with the changes AI is going to impose in the coming months and years.

But while substantial attention has been paid to attorney misuse of AI, very little has been directed toward its use by pro se litigants. As generative AI tools become more accessible, more litigants attempting to represent themselves are discovering that they can use them to draft letters, pleadings, discovery demands, and even conduct what passes for legal research. Unlike an attorney who faces professional repercussions for misusing AI, a pro se litigant faces no such consequences. The mix of pro se litigation and generative AI will likely marry the worst aspects of both worlds, and its increasing, unregulated use is likely to become increasingly problematic for pro se litigants themselves, attorneys appearing against them, and the courts.

The Competence Gap That AI Cannot Fill

The proper use of generative AI in a legal context requires more than familiarity with a chatbot interface. It requires a finely tuned understanding of how to prompt a model, what its outputs actually represent, and where its limitations and risks lie. An AI model does not inherently know the Rules of Court, or the difference between a summary judgment standard and a motion to dismiss standard. It does not inherently know the many procedural and evidentiary rules, jurisdictional distinctions, and local court rules and preferences that experienced attorneys handle on a day-to-day basis. It generates plausible-sounding text and superficially compelling arguments, neither of which is the same as accurate or correct.

For a pro se litigant, this gap is compounded rather than narrowed by the technology. The average self-represented party already has little to no familiarity with the relevant statutes, case law, court rules, and procedures that govern the dispute, and typically even lacks enough knowledge to prompt their respective LLMs to provide what they don’t know. AI does not inherently supply that missing foundation; it papers over it. A pro se litigant using AI to draft a complaint or a motion often produces something that reads fluently, uses the right vocabulary, and cites case law and statutes, some real and some invented, without any of the underlying legal judgment that should have shaped it. The result looks like lawyering. It is not.

The Practical Problems for the Courts

In recent months there has been a noticeable uptick in the number of pro se litigants relying on generative AI for pleadings and discovery demands, and the practical burden this places on the courts is significant, if not always visible from the bench in any single case. A judge or law clerk reviewing a pro se filing already has to do more independent work than with a filing from counsel, checking citations, construing pleadings liberally, and sorting substance from procedural defect. AI-generated filings raise the volume and the stakes of that work. A citation that looks legitimate on its face, complete with a case name, volume number, and reporter, may simply not exist, and it takes real time to run it down and confirm that it does not. Multiply that across an already overloaded docket and it becomes a genuine drain on judicial resources, not a hypothetical one.

There is also a fairness problem lurking underneath the efficiency problem. Courts are, and should be, more forgiving of pro se litigants’ procedural missteps than of counsel’s. That leniency was built for a world in which a pro se filing reflected a layperson’s own imperfect understanding of the law. It was not built for a world in which a pro se filing can be generated in minutes, dressed in the confident register of a legal brief, and submitted without any of the judgment, or accountability, that comes with a bar license. Courts will need to work out, case by case at first and eventually through rule or guidance, how much benefit of the doubt an AI-assisted pro se filing is entitled to when it is wrong.

The Practical Problems for Opposing Counsel

For the attorney on the other side, an AI-assisted pro se adversary is not necessarily an easier one. In some respects it is harder to litigate against. Pro se litigants historically have had trouble articulating claims, meeting pleading standards, and responding coherently to motion practice, and that has often created natural off-ramps, whether by way of a motion to dismiss, a failure to oppose, or simple attrition. Generative AI narrows those off-ramps. It can produce fluent oppositions, discovery responses, and even settlement correspondence that at least superficially meet the form, if not the substance, of what the rules require. Opposing counsel has to spend more time addressing filings that look substantive but may be built on hallucinated authority or a fundamental misunderstanding of the underlying law, and has to do so carefully, since directly attacking a pro se party over AI use risks looking like it is being used as a cudgel rather than a legitimate point of practice. There is also the practical question of proof. Absent an admission or an obvious fabricated citation, it is not always simple to establish that a filing was AI-generated at all, which limits how effectively, or even whether, an attorney can or should raise the issue with the court at all.

The Practical Problems for the Pro Se Litigant

The greatest cost of this trend, however, falls on the pro se litigants themselves, whether or not they recognize it. A litigant who relies on AI to draft a pleading or discovery response is trusting a tool that has no obligation to them, carries no malpractice exposure, and has no ability to exercise the kind of judgment that a matter often requires, such as when to settle, when to assert a defense, or how a particular fact pattern is likely to be received by a particular judge. When the output includes a fabricated case, the consequence falls on the litigant who filed it, not on the tool that generated it. Courts are not inclined to be sympathetic to a pro se party who submits fabricated authority merely because a chatbot produced it rather than a human. That litigant may face striking of the pleading, denial of relief, or in more serious cases, sanctions, all for relying on a tool that gave every outward appearance of trustworthiness.

There is also a slower, less visible cost. AI can make a pro se litigant feel more confident in a position than the underlying law actually supports, whether by validating a weak theory of the case or by failing to flag a defense, deadline, or procedural requirement that would have changed the litigant’s strategy entirely. A litigant who might otherwise have sought limited-scope representation, a legal aid clinic, or at minimum a consultation with counsel may instead proceed on the strength of an AI-generated document that reads persuasively but was never tested against the actual rules and requirements of the forum.

Where This Is Headed

This is not a passing curiosity. It is an early stage of a trend that will only accelerate as generative AI tools become more capable, more accessible, and more deeply integrated into everyday life. Courts will need to develop clear guidance on how AI-assisted pro se filings are to be treated, whether through certification requirements, standing orders, or amendments to court rules addressing disclosure of AI use. Bar associations and legal aid organizations have a role to play as well, in educating self-represented litigants on the real limitations of these tools before those litigants find out the hard way, in front of a judge, what those limitations are. In the meantime, practitioners should expect to see more of this, not less, and should be thinking now about how to address it efficiently, and without unnecessary hostility, when it shows up in their own cases.

Alan S. Golub, Esq. is a Principal of the firm and serves as the Managing Attorney of our Litigation and Business and Corporate Law Groups, where his practice focuses on complex commercial litigation, environmental disputes, business transactions, and intellectual property. Mr. Golub can be reached at agolub@fskslaw.com and at (973) 867-4538.

Robert E. Smithson, Jr., Esq. is a Senior Associate at the firm whose practice includes uncontested and contested foreclosure cases, municipal court defense, title curative matters, landlord-tenant disputes, and collection cases. Mr. Smithson can be reached at rsmithson@fskslaw.com and at (973) 867-4568