In 2023, a New York lawyer filed a federal brief citing Varghese v. China Southern Airlines Co., 925 F.3d 1339 (11th Cir. 2019). ChatGPT had supplied the case name, reporter citation, quotations, and judges. The opinion did not exist. A federal judge fined the lawyer and his firm $5,000, and Mata v. Avianca became a widely cited warning about AI-generated legal research.

A fabricated Tax Court case is the same object, and it breaks the same way.

The danger is that a fake can look exactly like a real citation. For a CPA or tax attorney about to put a case in a memo, that creates real exposure.

A case citation is a string the model finishes

Without a connected source, a language model can generate a case name and reporter citation from familiar patterns. The same problem produces invented IRC sections. A case citation is not useful until you can open the opinion behind it.

A real case can be just as dangerous

When Stanford researchers asked leading models specific, checkable questions about federal cases, the models fabricated an answer 69 to 88 percent of the time. But an invented case is only the most obvious failure. Case law adds two quieter ones.

CitationReal?What is actually there
Varghese v. China Southern Airlines Co., 925 F.3d 1339 (11th Cir. 2019)NoInvented by ChatGPT and filed in Mata v. Avianca; the opinion, the quotes, and the cite never existed
Renkemeyer, Campbell & Weaver, LLP v. Commissioner, 136 T.C. 137 (2011)YesHolds that partners’ shares from performing services are not shielded from self-employment tax by §1402(a)(13)
Soroban Capital Partners LP v. Commissioner, 161 T.C. 310 (2023)YesAdopts a functional-analysis test for the limited-partner exception

The first one does not exist. The other two do, and they are a click from the real text, if your tool can find them.

Two quieter failures involve real cases. A model may cite a case for a holding it never reached, or rely on an opinion later reversed or overruled. Confirming that the case exists is only the first check.

Tax Court opinions also carry different precedential weight. A regular “T.C.” opinion and a “T.C. Memo.” decision should not be treated as interchangeable merely because both appear in the same answer.

Before you cite

Until retrieval is part of your workflow, four checks stand between a fabricated case and a client deliverable.

  1. Open the citation. A real opinion should resolve to a real docket or reporter citation.

  2. Read the holding, not the model’s summary. A model can describe a real case as standing for something it never held. Pull the opinion and confirm it actually says what you are about to claim.

  3. Confirm it is still good law. A 2011 opinion can be undercut by 2014. Citation activity hints at currency; the subsequent history settles it.

  4. Keep drafting and authority in separate lanes. Let the model write the memo. Do not let it be the source of record for what the courts have held.

Cite from a corpus you can open

Retrieval from an index of opinions gives the model a real source to work from. You still need to confirm the holding and subsequent treatment.

TaxMCP’s Tax Court case law on Pro+ indexes more than 49,000 United States Tax Court opinions going back to 1942. Search by case name or citation to retrieve the opinion text and an openable source. Citation tools can also identify later Tax Court decisions that discuss the case.

The citation graph shows influence and currency, not validity. You still confirm good law before you stake a position. The tool gets you to the real opinion fast; your judgment takes over from there.

Start with the opinion

The practical rule is short: open the opinion, read the holding, and check its current treatment before the case reaches a memo or client deliverable. Let the model help with the draft after the authority is in hand.

Add Tax Court case law to your research with Pro+ →

Or start with the bigger question: can you trust AI for tax research at all?