How to Check a Generated Study Guide Before You Trust It
A generated guide reads fluently even where it is wrong. That is the difficulty: the sections are well formed and confident, so nothing in the presentation tells you which claims came from the passage and which came from somewhere else entirely.
The Study Guide Builder produces sections from a passage you supply. This article covers the check that should happen between generating and revising.
Quick answer
Trace each claim in the guide back to a specific sentence in the passage. Anything you cannot locate is either an addition from general knowledge or a distortion, and both should be removed or verified separately. Pay particular attention to numbers, names, definitions, and any claim stated more strongly in the guide than in the source. The check takes a few minutes and is the difference between a helpful guide and a confident set of errors.
Trace claims, do not read for sense
Reading the guide for sense will tell you it is coherent, which is not the question. The question is whether each statement is supported. Work claim by claim with the source open, and mark each one as found, not found, or found but different.
This is slow but bounded. A few hundred words of guide might contain twenty checkable claims, and most will trace directly. The ones that do not are the whole point of the exercise.
What to look for
- Numbers that do not appear in the passage, or appear with different units or rounding.
- Definitions phrased more broadly or narrowly than the source.
- Names and terms that are correct in general but not the ones the passage used.
- Claims stated as certain where the passage hedged, such as dropping a “usually” or “in most cases”.
- Causal links the passage did not make, connecting two things it mentioned separately.
The last one is the most common and the hardest to notice, because both halves of the claim are in the source. Only the connection is new.
Checking the structure as well
Sections that group unrelated material, or split one argument across two sections, are a structural fault rather than a factual one. They will not produce wrong answers, but they will make the guide harder to revise from, because no single section corresponds to a testable idea.
If the structure is wrong, regenerating with a narrower passage usually fixes it more reliably than editing the sections by hand.
After checking
Remove or correct anything that failed the trace. Keep the corrected guide alongside the source rather than replacing it, so you can check again later. Then convert the sections into questions you answer with the passage closed, which is the only way to know whether the material is actually available.
For the wider method, see How to Turn a Passage into Guide Sections You Can Revise From. For what makes a good test question, see What a Check Question Should Do.
Generate from one self-contained passage in Study Guide Builder and check every claim before you spend an evening revising from it.
Common questions
How do I check a generated guide quickly?
Trace each claim back to a specific sentence in the passage and mark it found, not found, or found but different. A few hundred words of guide usually contains about twenty checkable claims, and most trace directly. The ones that do not are the point.
What is the most common error in a generated guide?
A causal link the passage did not make, connecting two things it mentioned separately. This is hard to spot because both halves of the claim do appear in the source; only the connection is new.
What if the sections are grouped badly?
That is a structural fault rather than a factual one. It will not produce wrong answers, but no section will correspond to a testable idea. Regenerating with a narrower passage usually fixes it more reliably than editing the sections by hand.