In 2023, two attorneys submitted a brief to a federal judge in New York that cited six court cases as precedent — cases that did not exist. They had used ChatGPT to draft the brief, and ChatGPT had fabricated the citations with the same confident tone it uses when citing real ones. The attorneys were sanctioned, publicly embarrassed, and became a cautionary tale that has been repeated in legal circles ever since.
That story has scared many people away from AI legal tools entirely. That's the wrong lesson. The right lesson is that AI used carelessly in legal work causes real harm — but AI used correctly is a genuinely useful tool for people who otherwise have no affordable access to legal analysis. For pro se litigants, "used correctly" can make a meaningful difference in case outcomes.
The American legal system has a structural access problem. A civil litigation attorney typically charges $250–$600 per hour. A contested civil case — even a relatively simple one — can easily require 40–100 attorney hours. That means $10,000–$60,000 in legal fees for a dispute that may be worth far less, or for a plaintiff who has no way to pay regardless of merit.
The gap between those who can afford legal representation and those who cannot is not a small crack — it's a canyon. Legal aid organizations are chronically underfunded and can only serve a fraction of those who qualify. Law school clinics help, but they don't cover most case types. Court self-help centers can explain procedures but cannot give legal advice.
Into this gap, AI has arrived. The 2024–2026 generation of large language models — Claude, GPT-4, Gemini, and others — has legal knowledge that would have required a law degree to access five years ago. They can explain legal concepts, identify causes of action, analyze fact patterns against legal standards, and generate starting-point language for motions. For someone with no other access to legal analysis, this is not trivial.
The question is not whether to use AI. The question is how to use it without getting sanctioned or, worse, losing a case you had a right to win.
The dividing line is roughly this: AI is excellent at the conceptual layer of legal work — understanding, explaining, framing, organizing — and unreliable at the verification layer — confirming that specific cases exist, that a holding is correctly characterized, or that a rule currently reads the way it's described.
"Hallucination" is the term used when an AI model generates information that sounds authoritative but is factually incorrect. In legal work, hallucination most commonly takes the form of fabricated case citations — cases that don't exist, cases that exist but don't say what the AI claims, or real cases applied to a legal proposition they don't actually support.
This is not a bug that will be fixed next year. It is a fundamental characteristic of how large language models work. They generate text that is statistically probable given the patterns in their training data — and a confidently stated case citation looks statistically similar whether the case exists or not.
The consequences of citing a non-existent case to a federal judge range from having your motion denied and your credibility permanently damaged, to being sanctioned under Federal Rule of Civil Procedure 11 (which requires that every legal argument be "warranted by existing law or by a nonfrivolous argument" for extending or modifying existing law). Rule 11 sanctions can include monetary penalties and, in egregious cases, dismissal of your claims.
Courts across the country are rapidly developing rules on AI disclosure. The requirements vary significantly by jurisdiction, and this is an area where you must check your specific court before you file anything generated with AI assistance.
Several federal district courts have adopted standing orders requiring disclosure when AI tools were used to prepare court filings. The Northern District of Texas, the Southern District of New York, and others have issued orders requiring a certification that any AI-generated content has been reviewed and verified by the filing party. Some require specific disclosure language. Others simply encourage disclosure as a best practice.
State rules are evolving even faster, and in more varied directions. Some states have adopted mandatory disclosure requirements; others have issued advisory guidance; still others have said nothing yet. Check your state's court website and your specific local rules for any standing orders on AI usage.
When in doubt, disclose. Courts have generally reacted more favorably to parties who voluntarily disclose AI use and certify they verified the AI-generated content than to parties who use AI covertly and are later discovered. The act of disclosure signals professional responsibility and awareness of the tool's limitations — qualities that courts tend to view favorably in pro se litigants.
Used correctly, AI can save you dozens of hours of research time and help you produce more professionally organized legal arguments. Here is a workflow that extracts the benefit while managing the risk.
Start by explaining your facts to an AI tool and asking what legal theories might apply, what defenses exist, and what the elements of each claim or defense are. Use this output to understand the legal landscape — not to write your brief. Then go to primary sources (statutes, case law) to verify and build your actual arguments from there.
AI is excellent at helping you structure a motion — identifying the sections you need (introduction, statement of facts, legal standard, argument, conclusion), the logical flow of argument, and the questions you need to answer. Use it as a structural scaffold, then fill in the verified, sourced content yourself.
This bears repeating because the consequences of violating it are severe. Every case name, every statute citation, every regulatory reference — verify each one independently before including it in a court filing. This verification step is not optional.
Ask the AI to argue against your position. "What are the strongest arguments the opposing party could make against my motion?" This is one of the highest-value uses of AI in litigation preparation — it forces you to see your weaknesses before the judge does, and gives you the opportunity to address them in your filing.
If you use AI in preparing a filing and are later asked about it (by the court or opposing counsel), being able to produce your prompts and the AI outputs can demonstrate that you used the tool appropriately as a research starting point, not as a substitute for independent legal judgment.
ProperResponse's AI Legal Theory Generator is built around the principle that AI is most valuable as a structured research starting point — not as a source of final answers. When you describe your case facts, the generator uses Anthropic's Claude AI to:
Every output includes an explicit disclaimer: this is a research starting point, not legal advice. Case citations generated by the tool must be verified before use. The tool is designed to make you a better-informed pro se litigant — not to replace the legal judgment that only experience and a law license can provide.
Anthropic does not use content submitted to ProperResponse's Claude integration to train its AI models. Your case facts stay in your case. (See our AI Usage Policy.)
ProperResponse combines AI-powered legal theory generation with deadline tracking, document storage, and research tools for your judge and opposing counsel — all in one secure platform built for pro se litigants.
Start Your 3-Day Free Trial