More than three-quarters of tax court’s cases are pro se. These self-represented taxpayers are highly unlikely to be successful. Over half of these cases are dismissed during the pre-trial administrative process. If the case makes it to trial, only about one in ten of these taxpayers. In practicality, 99 percent of tax disputes resolve without a trial. Most non-attorney tax professionals use tax court to settle cases, not prepare them for trial without an attorney.

The AI Slop Problem

Now a new issue is rising in importance. Tax court is facing a rising tide of false filings based on AI slop. Attorneys can be sanctioned by the court, but pro se litigants are not subject to the same laws. The court knows that it must find a more effective disciplinary tool.

Tax Court has always dealt with frivolous filings. The difference now is volume and sophistication. A taxpayer can prompt a chatbot to “write a Tax Court petition challenging my deficiency notice” and receive something that looks like a real legal document, complete with case citations that don’t exist, legal theories that sound plausible but aren’t, and procedural language borrowed from actual filings. The court calls these “hallucinated” citations. Judges call them a waste of time.

Under Rule 33(b), the court can impose penalties up to $25,000 for frivolous or groundless proceedings. But enforcement against pro se litigants has historically been light. Judges understand that most of these taxpayers aren’t trying to game the system. They simply can not afford representation and don’t know what they don’t know. AI changes that calculus. When a taxpayer submits a 15-page petition packed with fabricated authorities, the court has to spend real resources determining that none of it is legitimate.

What the Court Is Likely to Do

We expect procedural changes ahead. The U.S. Tax Court has already signaled concern about AI-generated filings in other federal courts’ standing orders, and it’s only a matter of time before Tax Court follows with its own disclosure requirements or verification rules. Some possibilities:

1. A mandatory certification that AI tools were or were not used in preparing the petition.
2. A heightened screening process at the intake stage for petitions that trigger pattern recognition.
3. Expanded use of Rule 33 sanctions even against pro se filers who submit fabricated case law.

Most likely we will see all three.

What This Means for Tax Professionals

For a CPA, EA, or other non-attorney practitioner who uses Tax Court petitions as a settlement tool (that’s the correct use for most of us), this matters for two reasons.

First, the court’s patience with informal or imprecise filings is going to shrink. When judges are spending docket time sorting real petitions from AI-generated nonsense, they won’t extend the same grace to sloppy-but-legitimate filings that they once did. Your petitions need to be tighter than ever.

Second, your clients are going to start showing up with AI-drafted documents they found or generated themselves. They’ll ask you to “just file this.” Don’t. Review everything. If a client hands you a petition with case citations, verify every single one. Your Circular 230 obligations don’t disappear because the client wrote it.

The Big Picture

The real risk isn’t that AI will flood Tax Court with junk. The court will adapt. The real risk is that taxpayers will convince themselves they have a viable case because a chatbot told them so. They will be more inclined to bypass the administrative resolution process where 99 percent of disputes actually get settled. A taxpayer who files a premature or frivolous petition may lose their opportunity to resolve the matter through Appeals, where the odds are dramatically better and the cost is dramatically lower.

That’s the conversation I’m having with clients right now. Tax Court is a tool. Like any tool, it works when you use it correctly. Filing a petition to force IRS Appeals to the table is a legitimate strategy. Filing one because Claude wrote you a convincing brief is not.