We have all encountered AI-drafted or AI-assisted court filings and pleadings. No one enjoys reading or responding to them, for obvious reasons. An AI-drafted document is long. It is written with the unwavering confidence of someone who has never once experienced constructive doubt. It contains headings, transitions, abstractions, standards, caveats and rhetorical throat-clearing in industrial quantities. It may use the full page and word limits, often with innumerable nice-sounding adjectives. It gives the appearance of being polished. And, by the end, the reader is left with a hauntingly simple reaction: What, exactly, is this document saying?
This is becoming a more familiar problem in the age of generative AI. While some lawyers were already vague before the proliferation of AI, AI is exceptionally good at sounding analytical while offering majestic quantities of evasive prose. It can produce pages of legal-flavored language, complete with solemn transitions and respectable-sounding formulations, while avoiding any commitment to a clear proposition. It is the adversarial equivalent of being trapped at dinner next to someone who talks for 40 minutes about “the broader context” and never answers the question.
That creates a practical litigation problem. How do you respond to an argument that is not really an argument? How do you rebut something that sounds lawyerly, maybe even impressive, but remains so vague that any direct rebuttal risks dignifying it? And how do you handle the uncomfortable suspicion that the whole thing was drafted by a generative AI chatbot, then lightly dusted by a human hand and filed with the court?
You may choose to allege that your adversary has used AI, but that may not be advantageous. First, the mere use of AI as a drafting tool is not necessarily misconduct. And, even if it is misconduct, you may not be able to prove its use. It can be difficult to determine whether AI was used, to what extent it was used and whether its use was improper. Experienced litigators also recognize that maintaining a workable relationship with opposing counsel is often important to litigating a case effectively. For those reasons, think carefully before making any accusations.
Moreover, when responding to a vague brief—whether generated by a human or AI—the instinct may be to rebut every paragraph. That instinct should be resisted. A point-by-point response to a shapeless brief often does the opposing party a favor. It organizes the argument more clearly than the original party did. It translates abstractions into actual theories and gives the court the impression that there must have been something coherent there all along.
A better approach is to translate, expose and specify.
The first step is to translate the filing into plain English. After reading the submission, attempt to summarize the argument in three concise sentences. What, if anything, is the brief truly arguing? What authorities does it appear to rely on? Translating a brief means identifying the small number of propositions that the brief most plausibly advances. Those are now the issues. The haze has been converted into targets.
The second step is to expose what is missing. After sifting through the language and translating the brief, omissions tend to become easier to see. The document may never identify a governing test. It may recite multiple standards but never select one. It may cite authority without tying any case to a controlling proposition. Vagueness should be met with exposing identified omissions, not unhelpful outrage.
Finally, anchor every response in specifics. Vague arguments often survive because they remain abstract. The cure is specificity. For example, if the filing claims production was substantial, respond with the specific omitted custodians, date ranges and categories of documents. If the
filing insists there was no prejudice, specify the pending deadlines and consequences of the missing material. Concrete facts are the natural predator of synthetic vagueness.
The most effective opposition often follows a simple structure. State that the filing fails to articulate a clear theory. Identify the few arguments it appears to imply and answer those arguments with specific facts, elements and record citations. This approach preserves credibility, reduces noise and keeps the focus on what matters. Most important, it refuses to reward volume for its own sake.
The frustrating reality is that generative AI can produce endless litigation prose that sounds serious, lawyerly and even impressive, without ever clearly making a point. Effective legal writing still requires the basics: Identify the facts, create arguments and show why those facts support the arguments.
If a litigation filing says nothing, the response should not consist of an even lengthier document saying nothing in return. Instead, identify the absence of a real argument, reduce the dispute to the propositions that matter and make the court’s path through the haze as short as possible. In otherwords, when an adversary uses AI to cloud the path, the most effective response is to be the map.
Armin Ghiam is a partner in the intellectual property group of Hunton Andrews Kurth in the firm’s New York office. With an international background in law, finance and technology, he counsels clients on litigation, prosecution and transactions involving patents, trademarks, copyrights and trade secrets.
Kevin V. Small is a partner in Hunton’s insurance coverage practice in the firm’s New York office. A commercial litigator with prior experience as an insurance broker and advisor, his practice focuses on insurance coverage disputes and counseling on behalf of policyholders across a broad spectrum of policies.
Michelle Schoenwald is a summer associate in Hunton’s New York office.
Reprinted with permission from the August 10, 2026 issue of Litigation Daily. © 2026 ALM Media Properties, LLC. Further duplication without permission is prohibited. All rights reserved
