The stat sounds great. Does the study behind it survive?

A book quotes a study. An article says “research shows X reduces Y by 40 percent.” Before you build on it, the assistant traces the claim back through the actual paper, its methods, its data, and whether anyone has ever reproduced the result.

What it does

• Pins down exactly what’s being claimed: who, which paper, what number, over what period.
• Traces it to the original publication (title, authors, journal, year) and reads how the finding was actually measured.
• Checks whether the raw data is public, and whether an independent team has ever replicated the result.
• Searches for retractions, corrections, and published critiques of the paper.
• Returns a plain verdict: verified, partly verified, unverifiable, misattributed, or retracted, with the reasoning and honest caveats.

The reveal

Someone drops a compelling statistic into a meeting. You ask the assistant to check it. Twenty minutes later you learn the underlying paper is real but the citation oversold it: the study showed a correlation, not the cause everyone was repeating, and it was never replicated. You walk into the next conversation with the truth instead of the talking point.

What it doesn’t do

• It isn’t a takedown service. The point is rigor, not attack. If the study holds up, it says so plainly.
• It won’t call something “wrong” for lack of looking. “Unverifiable” means the evidence isn’t public, not that the search gave up.
• It doesn’t grade the science for you. It reports what the record shows and where the claim and the paper diverge.

Configuration

• The claim: paste the sentence, the stat, or the citation you want traced.
• How far to trace: just confirm the paper exists, or go all the way to raw data and replication.
• Output shape: a one-line verdict, or the full trace with sources.

Triggers

On-demand. You bring a claim or a citation and the trace runs in the background.