A verification layer is a smoke alarm, not a building code.

Citation verification is a legitimate control. A firm-wide checker that catches a fabricated case before it reaches a judge is exactly the right response to the filings that have drawn sanctions. Every building should have a smoke alarm, and a building without one is negligent.

But an alarm tells you something is already burning. It builds nothing, and it structurally cannot detect several of the ways briefs actually fail. A verifier reads the document, so it can check only the citations that made it onto the page. It passes the brief whose controlling adverse case was never retrieved, the brief that applies the wrong procedural standard with perfectly accurate quotations, the uncited assertion that offers nothing to check, and the valid authority marshaled for the wrong procedural instrument.

Even on its own ground, detection scales badly. With 35 cited propositions and a 2 percent miss rate, fewer than half of briefs pass clean. Used as an inline gate, it invites Goodhart's law: the cheapest way to pass a citation check is to cite less. And human factors research since Lisanne Bainbridge's "Ironies of Automation" (1983) predicts the worst effect, that people working alongside a reliable partial control become worse at catching what it misses. A stamp that says "checked" erodes the scrutiny that used to cover the gaps.

The building code is architecture: a system in which the model never originates a citation, every authority traces to a retrieved record, adverse authority is searched for deliberately, and the process leaves a record of what was searched, retrieved, rejected and reviewed. Keep the alarm. Do not mistake it for the building.