A clean summary feels like research because it is clean. That is the whole trap.
The bullets line up, nothing contradicts itself, and it reads like a finished conclusion. But polish is a writing quality. It says nothing about whether the figures are right or whether anyone checked. Smooth and verified are not the same thing.
A summary tells you what a document said. Research proves it, ties it to a source, and tries to break it. AI is excellent at the first job and does not do the other three on its own.
The three steps a summary skips:
1. Source tie-out: every figure traces to a filing, a page, and a date, re-derived from the primary document. No trace, not usable. 2. Verification: open the actual filing or transcript and confirm the number and the caveat independently. Summaries quietly drop footnotes, and the footnote is often the point. 3. Challenge: ask what would make this wrong, go look for the disconfirming line, and force a positive / negative / unclear call. Unclear goes on the follow-up list.
The failure mode is simple. You read the clean summary, it agrees with what you already believed, you stop, and an unverified number rides into your thesis. Clean is exactly how it gets past you.
Disciplined version: AI builds the summary fast, you do the tie-out, the verification, and the challenge. AI summarizes. You research.
Save this one and run it on your next AI output.
Educational content only. Not investment advice, and not a recommendation to buy, sell, or hold any security. Wall Street Prompt. Always verify against the primary source.