Citation Verification

Accuracy has always been fundamental and essential to legal work product. It is non-negotiable. Showalter PLLC’s lawyers have always taken pride in the diligence and care they bring to their work. Lawyers have sometimes gotten into trouble when using AI because of failures of diligence and care. They have been sloppy. Showalter PLLC’s lawyers are not sloppy. Showalter PLLC’s lawyers find other lawyers’ mistakes.

Every citation in every Showalter PLLC filing is verified against primary legal databases. No citation is included on the word of a model.

Two layers of checking. The first layer is deterministic: each case name, court, and reporter volume is matched against legal databases, and the match succeeds or fails. That layer cannot hallucinate, because it looks up a record. The second layer uses a model to test whether each authority supports the proposition it is cited for.

Proven against opposing counsel. This summer, our cite-checking skill flagged in three minutes that opposing counsel in a Ninth Circuit appeal had rested a central argument on a hallucinated quotation. We raised it in our reply brief, and opposing counsel moved for leave to file an amended brief.

Better than the old way. We find fewer ordinary citation mistakes after our AI cite check than we did with the human paralegals at large firms. Run on law review articles that had already been through human cite-checking, it found a long list of errors.

Michael Showalter explains why in a public comment on the Tenth Circuit’s proposed rule on generative AI.