Chapter 14Check before you trust
How to check
Match the check to the claim: redo the math, open the source, search the original, look it up elsewhere, or ask a person who knows. Check more when more is at stake.
Read sideways. In a study of professional fact-checkers, Ph.D. historians and Stanford students judging websites, the fact-checkers did best: they “read laterally, leaving a site after a quick scan and opening up new browser tabs” to see what others said about it, and reached better-supported conclusions in a fraction of the time.2 The same habit works on an AI tool’s answer: check it somewhere else.
Match the effort to the stakes. A dinner suggestion needs no checking; a number going to your board, a quote in a report or anything about a person’s health or job needs all of it. Chapter 12’s sensitivity levels are a good guide.
How this book did it. Every claim was graded before it shipped, and grade D, seen only in a snippet or summary, never ships. For the newest chapters, the full text of every source was obtained and read (the papers, the court order, the laws themselves), every quote was matched word for word, and each claim was checked three times: against an independent record of the source, in the source itself, and by a second reader who opened the source cold. See Appendix E.
Check in
Try answering in your own words, then open each one to compare.
Why does fluent text deserve a second look?
A model writes what sounds right, and people working with automated aids can trust them too much. Fluent isn’t the same as checked.
What is the hardest kind of claim to check?
A claim that something is missing. Search the original yourself.
Once it’s checked, it’s nearly ready to go. Next: what to do before you share it. On to Chapter 15: Before you share →
Notes (2)
- 1U.S. District Court for the Southern District of New York (P. Kevin Castel, U.S. District Judge), “Mata v. Avianca, Inc., No. 22-cv-1461 (PKC): Opinion and Order on Sanctions (ECF No. 54),” No. 22-cv-1461 (PKC), ECF No. 54, June 22, 2023, https://www.courtlistener.com/docket/63107798/mata-v-avianca-inc/ (accessed October 27, 2026). Cited at p. 1.
- 2Sam Wineburg and Sarah McGrew, “Lateral Reading and the Nature of Expertise: Reading Less and Learning More When Evaluating Digital Information,” Teachers College Record 121, no. 11 (November 2019): 1–40, https://doi.org/10.1177/016146811912101102. Cited at abstract.
Chapter 14 · page 2 of 2
Turn pages with the ← and → keys.