Understand and Apply sit next to each other in Bloom’s taxonomy, and they’re the two levels that get blurred most. The objective says learners will apply a policy. The quiz asks which statement describes it. Everything looks finished, and learners never practice the thing the course promised.
It’s a hard line to see, even for software built to see it. Two studies that trained AI to sort content by Bloom’s level found the same thing: the middle levels, especially Understand and its neighbors, get mixed up more than any others.1 If a model built for that one job struggles, a quiz written at 5pm on deadline will too.
What each level asks learners to do
| Understand | Apply | |
|---|---|---|
| The learner… | Shows they grasp what something means | Uses it to handle a new, realistic case |
| Typical verbs | Explain, summarize, classify, compare, give an example | Use, carry out, solve, demonstrate, choose the action |
| The answer is… | A restatement or a sorted example | A decision or an action in a situation they haven’t seen |
| Usually looks like | “Which statement best describes…” | “Here’s what happened. What do you do first?” |
The 30-second check
- Circle the verb in the objective. That’s the level you promised.
- Look at what the learner actually does to answer the question. Ignore the wording of the stem. What work does answering it take?
- If they’re picking a definition or a restatement, it’s Understand, or Remember, whatever the objective says.
- If they have to act on a case they haven’t seen, it’s Apply. The case has to be new: a scenario copied from the lesson is closer to recall.
Side by side
| Objective (Apply) | Question at Understand | Question at Apply |
|---|---|---|
| Choose the correct action when an email looks suspicious. | What does the word “phishing” mean? | An email from “HR” asks you to update your benefits. The sender is hr@company-external.com. What do you do first? |
| Submit a travel expense report that meets the policy. | Which statement describes the per diem rule? | Here are three receipts from a two-day trip. Which can be claimed, and for how much? |
| Give corrective feedback using the SBI model. | What do the letters in SBI stand for? | An employee missed two deadlines this month. Which of these openings follows SBI? |
The tells that a question sits below Apply
- The stem starts with “Which statement describes…”, “What is…”, or “Which of the following is true about…”
- The right answer appears word for word in the lesson.
- There’s no situation: no person, no constraint, nothing to decide.
- A learner could answer it without ever having tried the task.
Rewriting a recall question into Apply
Objective: Choose the correct action when an email looks suspicious.
Q3: What does the word “phishing” mean?
Q3: You receive an email from “HR” with a link to update your benefits. The sender address is hr@company-external.com. What is your first action? A) Click the link. B) Report the email. C) Delete it.
Three moves made the difference: a realistic situation, a detail the learner has to notice (the outside sender), and options that are all things a real person might do. The wrong options aren’t silly, so picking the right one takes the skill.
Why it slips, and when to look
A definition question takes a minute to write. A good scenario takes ten. Under deadline, and in AI-drafted courses, quizzes drift toward whatever is fastest to produce, and that’s usually the level below the objective. Check for it when the quiz is first drafted, not at the final review, when rewriting means reworking the practice too. The five-minute alignment check catches the missing links; this check catches the links that are there but at the wrong level. For the verbs at every level, see Bloom’s taxonomy verbs, and for AI drafts in general, how to review an AI-generated course.
Checking it across a whole course
For one quiz, this check takes a few minutes by hand. For a full course, Pedagrade reads every question against its objective, flags the ones a level too low, quotes the question, and writes the Apply-level rewrite. You still decide on each one.
1. Sources: eXplainable AI Framework for Automated Lesson Plan Generation and Alignment with Bloom’s Taxonomy, Computers (MDPI), 2025, doi.org/10.3390/computers14110494; and ARCHED, arXiv 2503.08931. Both report AI classifiers confusing neighboring Bloom’s levels most often in the middle of the taxonomy.
See what it finds in your course.
Paste a course or upload the file. In about a minute you get a scored report with every finding cited and the fix written for you. Free for a week, no card needed.
Start your free review