How to automatically grade online quizzes.

Which formats grade themselves

Single choice, multiple choice, and fill-in-the-blank questions are graded automatically the moment a participant submits. Free text answers are collected for you to read and grade yourself. Planning a quiz with this in mind lets you decide how much manual marking you are willing to do before you share it.

A common balance is a mostly auto-graded quiz with one or two free text questions where you genuinely need to see reasoning.

The answer key is everything

Automatic grading is only as accurate as the correct answers you set. A single wrong key marks every participant wrong on that question, and the error may go unnoticed until someone complains. Before sharing, check every correct answer, and if possible take the quiz yourself or ask a colleague to take it.

Pay extra attention to AI-drafted and imported questions, since their answers were not typed by you.

Fill-in-the-blank needs care

Fill-in-the-blank questions grade on an exact match with the answer you set. That makes them strict: a participant who writes a correct answer in a slightly different form may be marked wrong. Reduce this risk by asking for a single word or number, stating the expected format in the question, such as 'answer in whole numbers', and avoiding answers with common spelling variations.

Setting points

Points can differ by question, which lets you weight harder or more important questions more heavily. For example, a two-step application question might be worth more than a simple recall question. Show the points in the question text if participants should know, so they can decide how to spend their time.

Grading free text efficiently

For free text answers, write a short model answer and a few criteria before reading any responses. Grading against criteria is faster and more consistent than deciding each time. Read all answers to one question before moving to the next, which makes it easier to compare responses fairly.

Keep free text questions to a minimum on large groups, since each one adds reading time per participant.

Check results before trusting them

After responses arrive, open Question Performance. A question that most participants got wrong may reveal a genuine learning gap, or it may reveal a mistake in the question or answer key. Reread any surprising question before acting on it. A few minutes of checking prevents unfair scores from being shared.

What participants see

Participants see a completion screen when they finish, and a leaderboard if you enabled one. For practice quizzes, the leaderboard adds motivation. For graded tests where fairness matters, consider leaving it off and sharing results in your own way after everyone has taken the test.

Using the time you save

Automatic grading removes hours of marking. The best use of that time is acting on the results: re-teaching the topics that caused problems, talking to participants who are struggling, and improving weak questions. Bookmarking missed questions into Revision and creating a re-test later closes the loop on learning.

Exporting grades

Scores, correct counts, and time taken are available in the quiz detail view and can be exported as CSV for spreadsheets or PDF for sharing. Individual responses can be opened for a question-by-question breakdown, which is useful when a participant asks why they lost marks.

Example: grading a class of forty

A teacher gives forty students a twenty-question test with eighteen auto-graded questions and two free text questions. The eighteen are scored the moment each student submits. She reads only the eighty free text answers against a short model answer, instead of marking eight hundred answers by hand, and finishes grading in well under an hour.

Quick setup checklist

Every correct answer checked. Points set per question. Fill-in-the-blank answers short and unambiguous. Free text limited to where reasoning matters. Quiz taken once yourself to confirm scoring before sharing.