Writing Distractors That Make Multiple-Choice Questions Worth Asking

The quality of a multiple-choice item is decided by the wrong answers, not the right one. Ask AI simply to "make a four-option question" and only the correct answer is plausible, while the remaining options are filler anyone can see is wrong. That gives you a question students get right by guessing.

What makes a good distractor

A distractor with discriminating power aims squarely at a mistake students commonly make. You have to ask AI for the following.

• Built on misconceptions: say it explicitly, "make the wrong answers out of mistakes students often make."
• Plausible: match the length and format of the correct answer so nobody can guess it by length alone.
• One clear error each: for every distractor, you should be able to explain in one sentence why it is wrong. If it is fuzzy, you get arguments about multiple correct answers.

In one real case, a middle school English teacher writing an item on tense asked specifically for "options a student who confuses the present perfect with the simple past would pick," and got an item that could not be solved by rote memory.

What to filter out after writing

1. Check for giveaways: look for words in the stem that hint at the answer.
2. Confirm mutual exclusivity: check for the possibility of two options both being correct.
3. Beware "all"/"none" options: AI overuses "all of the above" and "none of the above," so ask whether they are truly needed.
4. Write distractor explanations: produce a short explanation for each wrong answer and use it in feedback after grading.

A good item separates the students who know from the students who don't. What draws that line is a well-designed set of distractors.

Refining items with data after the test

Writing the item once is not the end. Look at how students responded after the test and the item's quality shows.

• Check the percent correct: above 95 percent and it discriminates nothing; below 20 percent and it may be too hard or flawed. The workable range is generally 40 to 80 percent.
• Look for clustering on one distractor: if students piled onto a particular wrong answer, that is a signal the misconception is widespread. Address it in the next lesson.
• Reversed performance at the top: an item that the strongest students got wrong more often means the trap is overdone or the wording is ambiguous.
• Blank response rate: an item nobody answered was most likely unclear in its instructions.

One teacher tabulated percent correct after each test, raised the difficulty of items that everyone got right, and reworded the ones the top students missed, lifting the quality of the item bank year after year. Good items are not finished in one pass; they are refined with data.

Key takeaways

Multiple-choice quality comes from attractive distractors aimed at misconceptions. Ask AI for wrong answers grounded in common mistakes, then check for giveaways, multiple correct answers, and explanations, and the discriminating power rises substantially. After the test, refine items with data such as percent correct and clustering on distractors, and you build a better item bank every year. Start with the items for a single unit, put the work into the distractors, and cut down on questions students get right by guessing. Even a little data, gathered consistently, sharpens your instincts as an item writer noticeably.