Module 2: Citators & Validating Authority

A citator — Shepard’s Citations or KeyCite are the two dominant commercial versions — tracks every later case, statute, or secondary source that has cited a given authority, and categorizes how that authority was treated: affirmed, reversed, questioned, distinguished, criticized, or simply cited without comment. It answers a question a plain keyword search never can: is this authority still valid?

A case appearing in your search results tells you nothing about its current validity. Search relevance and subsequent judicial treatment are entirely different things — a case can rank highly in a keyword search and still have been reversed the year after it was decided.

Citator checking predates computers by more than a century. Frank Shepard created the original citator system in 1873, and it started as something genuinely physical: small, gummed paper labels — “Shepard’s Adhesive Annotations” — that subscribing lawyers literally pasted directly onto the margins of their own law books, next to each case, listing every later citing case a researcher needed to check. That manual, sticker-based origin is the real reason “Shepardizing” became, and remains, a genuine verb in legal practice, still used today even by researchers who exclusively use KeyCite or another electronic citator and have never touched a physical Shepard’s volume. The underlying task Shepard designed his stickers to perform — systematically tracking every later citation to a given authority — is identical to what a modern electronic citator does; only the mechanism changed.

Citators summarize subsequent treatment with a signal — a color or symbol meant to prompt further reading, never to substitute for it. A signal alone doesn’t tell you whether the specific point you’re relying on is actually affected.

This is the single most important nuance in citator work: a case can be overruled on one holding while remaining perfectly good law on another, within the same opinion. A negative flag means “go read what actually happened” — not “discard this case entirely.” Conversely, a case with some negative history elsewhere in the opinion may still be exactly the authority you need, if the specific point you’re relying on was never touched.

Payne v. Tennessee, 501 U.S. 808 (1991), shows the specific-point distinction operating at the Supreme Court level. Two earlier decisions, Booth v. Maryland (1987) and South Carolina v. Gathers (1989), had held that victim-impact evidence — testimony about the harm a crime caused the victim’s family — was categorically inadmissible during a capital sentencing proceeding. In Payne, the Supreme Court expressly overruled Booth and Gathers, but only on that specific victim-impact-evidence holding; nothing else those two opinions addressed was disturbed. A researcher running Booth or Gathers through a citator today would correctly see a strong negative signal — but that signal is doing real, specific work: it flags the victim-impact holding as overruled, not the entirety of either opinion. Payne is a clean, real-world illustration of exactly this section’s warning: seeing a negative citator flag and concluding “this case is dead, don’t cite it for anything” is itself a research error, just a different one than failing to check the flag at all.

Citator discipline isn’t limited to case law. A statute can be amended, repealed, or held unconstitutional after the fact — relying on outdated statutory text, or on a case interpreting a version of a statute that no longer exists, is exactly the kind of error a citator check is meant to catch.

United States v. Windsor, 570 U.S. 744 (2013), is a real, concrete illustration of exactly this trap. Section 3 of the federal Defense of Marriage Act had defined “marriage” and “spouse” for all federal-law purposes as limited to opposite-sex couples, and for roughly seventeen years, courts across the country issued opinions applying that definition to federal tax, benefits, and immigration questions. The Supreme Court struck down §3 as unconstitutional in Windsor. A researcher today pulling an older case that applied §3’s now-invalidated definition — without running a citator check on the underlying statute itself, not just the case — could easily miss that the entire statutory basis for that case’s reasoning no longer exists, even if the case itself was never technically overruled by name. Windsor is exactly why this section insists statutory validation is a genuinely separate, necessary step from case validation: a case can remain formally uncontradicted by any later opinion and still be built entirely on statutory ground that has since been removed.

Citing invalid authority isn’t just a research shortcoming — it can mislead a tribunal and expose the attorney to real professional consequences. In Mata v. Avianca, Inc. (S.D.N.Y. 2023), attorneys submitted a brief citing six cases that turned out not to exist at all — generated by an AI tool and never verified before filing. The court sanctioned the attorneys and required them to notify every judge falsely cited in their papers, turning an avoidable verification failure into a widely publicized professional embarrassment.

The lesson generalizes well beyond AI-generated content. Whether a citation came from a careless keyword search, an outdated research memo reused without rechecking, or a tool that fabricated it outright, the underlying discipline is the same: nothing goes into a filing without being independently validated first.

This obligation is written directly into procedural law, not just professional custom. Federal Rule of Civil Procedure 11(b)(2) requires every attorney signing a filing to certify, on personal knowledge, that its legal contentions are “warranted by existing law” — a certification that is simply false if the authority cited was never actually validated and turns out to be overruled, repealed, or entirely fabricated. Most state courts impose an equivalent certification requirement under their own analogous rules. Citator discipline, in other words, isn’t merely good practice recommended by legal-writing instructors — every signed filing already contains an implicit, legally binding promise that this exact validation step was actually performed, which is precisely why courts like the one in Mata v. Avianca treat skipping it as sanctionable misconduct rather than a forgivable oversight.