Students & Study

Verified against ChatGPT · 2026-08-11

Generate comprehension questions that actually test different levels of understanding, not ten variations on 'what happened'

Builds a set of reading comprehension questions across distinct cognitive levels — literal recall, inference, and evaluation — from a specific passage, so self-testing catches the gap between remembering a passage and actually understanding it.

ChatGPT (GPT-5.1)3 fillable variables

The prompt

Ready to copy — highlighted parts are example details you can swap.

Generate reading comprehension questions from the passage below, spread across genuinely different levels of understanding rather than ten questions that all just test recall of what happened.

PASSAGE
A 600-word excerpt from a science journalism article on why octopuses are considered unusually intelligent despite short lifespans.

WHAT THIS IS FOR
Practicing for a standardized reading comprehension section that mixes detail, inference, and author's-purpose questions.

QUESTION LEVELS AND HOW MANY OF EACH
3 literal, 4 inferential, 2 evaluative.

Build questions at three distinct levels, matching the counts I specified: (1) literal — answerable by pointing to one specific sentence or detail in the passage; (2) inferential — requires connecting two or more parts of the passage that aren't stated together, or reading between the lines of something implied but not said outright; (3) evaluative — requires a judgment about the passage itself, like identifying an author's assumption, a weakness in an argument, or what evidence would change the conclusion. Label each question with its level so I know what kind of thinking it's testing. For every question, write an answer key entry that doesn't just give the answer but names exactly where in the passage it comes from (a quote or paraphrase) so I can check my reasoning, not just my final answer. For inferential and evaluative questions specifically, the answer key must explain the reasoning chain, not just assert the conclusion — if a question requires connecting two facts, name both facts and how they connect.

Do not write an evaluative question that actually only requires locating information (a common failure mode where a question is labeled "analyze" but is answerable straight from the text) — if you can't construct a genuine evaluative question from this passage, say so and tell me why rather than mislabeling a recall question to hit the count.

Output as three labeled sections (Literal, Inferential, Evaluative), each with numbered questions, followed by a separate answer key section in the same order.

Customize

Optional — swap in your own details for the highlighted parts above.

Why this works

Without an explicit level distinction, GPT-5.1's default question-generation behavior clusters heavily around literal recall, because recall questions are the easiest to generate reliably correct answer keys for and the model has no signal that you want the harder, more ambiguous inferential and evaluative categories represented — naming the three levels and forcing a labeled split makes the model allocate effort to the categories it would otherwise under-produce. The explicit ban on mislabeling a recall question as "evaluative" targets a specific and common failure mode in AI-generated study questions: a question that sounds analytical ("analyze why the author chose this structure") but is actually answerable by finding one sentence that states the reason directly is a fake-inference question, and self-testing against it teaches false confidence because getting it right doesn't confirm you can actually make inferences, only that you can locate text. Requiring the answer key to show the reasoning chain rather than just the final answer is what makes this usable for self-study specifically: a bare answer key lets you confirm you got a question right or wrong but not why, whereas naming both facts an inferential question connects and how they connect lets you diagnose whether a wrong answer came from missing a detail versus failing to connect two details you did notice — those are different remediation paths, and collapsing them into one letter grade of correct/incorrect wastes the diagnostic value the exercise could have had.

What you get back

Inferential Q2: Why might the article's structure devote more space to octopus cognition than to their lifespan, despite opening with the lifespan paradox? Answer: The article states octopuses live only 1-2 years (paragraph 1) and separately details extensive problem-solving research (paragraphs 3-5); connecting these suggests the author frames the short lifespan as the surprising contrast that motivates deeper explanation of the cognition research, using structure to build toward the paradox's resolution rather than treating both facts as equally weighted.

Verified against

ChatGPT GPT-5.1 · 2026-08-11

Changelog

  • 2026-08-11 Initial publish, verified against ChatGPT GPT-5.1.

Need this built into your business?

If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.

EXPLORE WHAT SCULT BUILDS
All Students & Study prompts

Check your AI visibility

One URL in, a 0–100 score and the exact fixes out.

RUN THE CHECK

Browse all the tools

15 tools across six categories
13 of them never send your data anywhere

Free · No signup · No trial clock

SEE THE DIRECTORY