AI essay grading in QuizMagic reads each student’s full written response and evaluates it against the grading criteria you define, producing a suggested score and criterion-by-criterion feedback for every submission automatically. Rather than replacing your judgment, this feature within QuizMagic’s AI quiz generator gives you a detailed first-pass evaluation that you review and finalize in a fraction of the time manual grading takes.
This article covers how to create an essay quiz, how to configure grading criteria in Smart Sharing, how to review AI-generated scores, how tokens work, and how to get the most accurate and useful results from the AI grader.
Premium feature. AI essay grading is available to Premium subscribers only. Free users and premium trial users cannot access AI essay grading because essay grading tokens are not included in the free premium trial. To unlock this feature, upgrade at quizmagic.io/pricing.
How AI Essay Grading Works
Before walking through the setup, it helps to understand what the AI actually does when it evaluates a student’s response.
When you use QuizMagic for auto-grading quizzes, the system picks up and sends the students’ full written responses to the AI alongside the grading criteria and percentage weights you configured before the quiz was shared. The AI reads the entire response in context, not sentence by sentence, and evaluates how well it satisfies each active criterion. It then assigns a score per criterion, calculates a weighted composite score, and generates a narrative feedback summary explaining what the student did well and where their response fell short.
This process happens in seconds per submission. A batch of 35 essays generates 35 preliminary evaluations faster than you could finish reading a single response manually.
Importantly, the AI evaluation is always a first pass. Nothing is released until you grade student responses by reviewing every submission, adjusting scores where needed, and giving final approval. You retain complete authority over every grade.
Note: AI essay grading works differently from objective question grading. Multiple Choice, True/False, and Fill-in-the-Blanks questions are graded by matching responses against a correct answer. AI Essay grader requires the AI to read and reason about a student’s full written argument, which is a fundamentally different process. For this reason, essay quizzes are a separate quiz type and cannot currently be combined with objective question types in the same assessment.
Understanding Essay Grading Tokens
AI essay grading uses a token system to manage the computational cost of reading and evaluating full written responses.
| Token detail | Value |
|---|---|
| Tokens included with Premium | 5 tokens per month |
| Essays covered per token | 1 token = max of 100 responses (students x questions) 1 token can grade a 4-item essay quiz for 25 students |
| Monthly essay grading capacity (included) | Up to 500 essay responses |
| Additional tokens | $1.00 per token (covers 100 more items) |
| Token availability on Free plan | Not available |
| Token availability on Premium trial | Not available – tokens are not included in the free trial |
How token usage works
QuizMagic consumes a token when you initiate AI grading for a batch of essay submissions. A single token covers up to 100 responses in one batch. A 2-item essay quiz session for 35 students (70 responses) uses one token. A 3-item essay quiz session for 35 students (105 responses) uses two tokens.
Premium gives you 5 essay-grading tokens per month. Monthly tokens reset (do not roll over) on your subscription renewal date. Purchased tokens never expire, and QuizMagic only uses them after you exhaust your 5 free monthly tokens. You can purchase additional tokens at any time from your account settings at $1.00 per token.
Step 1: Create an Essay Quiz
AI essay grading requires a quiz saved with the Essay question type. If you have not yet created your essay quiz, start here.
Go to the QuizMagic dashboard and choose your source: type a topic, paste text, upload a file, or use a YouTube link. Under Quiz Type, select Essay. Configure the number of questions, difficulty, and cognitive framework settings, then click Generate Quiz.
The AI produces essay prompts based on your source material. Review and edit each prompt in the quiz editor before saving. You can rewrite prompts, add your own manually, or delete any that do not suit the assessment.
When you are satisfied with the prompts, save the quiz or click Share right away. If you chose save, it appears in your Saved Quizzes dashboard, ready for sharing configuration.
Writing good essay prompts
The quality of AI grading correlates directly with the clarity and specificity of the essay prompt. Vague prompts produce vague responses that are harder to evaluate consistently. Specific prompts produce responses that are more clearly aligned, or misaligned, with the learning objective, making AI evaluation more accurate.
Less effective: “Discuss photosynthesis.”
More effective: “Explain the relationship between the light-dependent reactions and the Calvin cycle in photosynthesis. In your response, describe what each stage produces and why both stages are necessary for glucose synthesis.”
The second prompt tells the AI exactly what to look for in a strong response, which produces more precise criterion-by-criterion feedback for students.
Step 2: Open Sharing Options and Select Smart Sharing
From your Saved Quizzes dashboard, locate your essay quiz and click Sharing Options. This opens the Quiz Sharing Manager. If you click Share directly from the Generated Quiz section, you’ll be asked to enter the quiz name. Once done, the Quiz Sharing Manager opens with Smart Sharing automatically selected (Simple Sharing is not available for essay-type quizzes).
Smart Sharing is the only sharing mode that supports AI essay grading. Simple Sharing does not include the grading criteria configuration panel or the AI grading engine.
Give your session a name, for example, “Argumentative Essay – Class A – Week 6”, so you can identify it easily in your results dashboard later.
Step 3: Configure Session Settings
Before reaching the grading criteria, configure the core session settings that apply to all Smart Sharing sessions.
Time Limit (optional but recommended). Toggle the time limit on and set a session duration in minutes. For essay quizzes, allow enough time for students to write a complete response, typically 20 to 45 minutes for a single essay prompt and up to 90 minutes for longer multi-prompt assessments.
Allow Review. For essay quizzes, leaving this off is common practice since you still have to run the AI Essay Grading once all submissions are in. For non-essay quizzes like MCQ or True/False, this toggle controls whether students see their scores and feedback after submitting.
Anti-Cheating. Toggle anti-cheating on to activate tab-switch detection, copy/paste blocking, right-click blocking, and DevTools monitoring during the essay session.
Step 4: Configure Grading Criteria
This is the most important step in the AI essay grading setup. The grading criteria you configure here tell the AI exactly what a strong response looks like for this specific assignment. The AI evaluates every submission against these criteria, so the precision of your configuration directly determines the usefulness of the feedback students receive.
The Grading Criteria panel appears below the Anti-Cheating toggle in the Smart Sharing configuration. It shows six criteria, each with a checkbox and a percentage weight field. Your active criteria weights must total exactly 100 percent before you can create the session.
The six grading criteria
| Criterion | What the AI evaluates | Best used for |
|---|---|---|
| Relevance | How directly and completely the student’s response addresses the actual question asked | All essay types – this is almost always active |
| Grammar and Spelling | Correctness of language mechanics, sentence structure, and spelling | Writing-focused courses, ELA, language assessments |
| Coherence | Logical flow, organizational structure, and clarity of argument | Argumentative essays, analytical writing |
| Depth of Analysis | Thoroughness of reasoning, quality of insight, and evidence of genuine engagement with the content | Higher-level courses, Bloom’s Analyzing and Evaluating levels |
| Factual Accuracy | Whether the specific claims, facts, data, and references in the response are correct | Science, History, Law, any content-specific assessment |
| Originality | Independent thinking, unique perspectives, and insights that go beyond restating familiar material | Advanced coursework, research-based writing, creative analysis |
How to assign weights
Check the criteria relevant to this assignment and enter a percentage in the weight field next to each. The total must reach exactly 100 percent. The panel shows a running total and confirms with a green checkmark when you reach 100.
Uncheck any criterion you do not want included. Unchecked criteria receive zero weight and are excluded from scoring and feedback entirely.
Important: The weights you assign directly shape what the AI rewards in student responses. A criterion weighted at 40 percent has four times the influence on the final score as one weighted at 10 percent. Assign weights that honestly reflect what matters most for this specific assignment.
Weighting examples by assignment type
Argumentative essay (History, Social Studies): Relevance 35%, Depth of Analysis 35%, Coherence 20%, Factual Accuracy 10%
This configuration rewards clear engagement with the prompt, substantive reasoning, organized argument structure, and factual accuracy, in that order of priority.
Literary analysis (English Language Arts): Relevance 30%, Depth of Analysis 30%, Coherence 25%, Originality 15%
This weighs analytical insight and original interpretation alongside structural clarity, reflecting what literary analysis teachers most commonly assess.
Science lab report summary: Factual Accuracy 40%, Coherence 30%, Relevance 20%, Depth of Analysis 10%
Science writing prioritizes accuracy above all else. This configuration penalizes factual errors heavily while still rewarding organized, relevant responses.
First-year composition (writing fundamentals): Relevance 25%, Grammar and Spelling 25%, Coherence 25%, Depth of Analysis 25%
For courses where writing mechanics are an explicit learning objective, equal weighting across four criteria reflects a balanced evaluation of both form and content.
Corporate training scenario response: Relevance 50%, Depth of Analysis 30%, Coherence 20%
For professional training assessments, the primary question is whether the employee understood what the scenario required and can articulate the correct response. Grammar and spelling are largely irrelevant in this context.
Step 5: Set Points Per Question
Below the Grading Criteria panel, the Points Per Question section lets you assign a raw point value to each essay prompt individually. For example, a quiz with four essay questions might assign 10 points each for a total of 40 points, or weight a longer analysis question at 20 points and shorter response questions at 10 points each.
The AI scores each response as a percentage of the criteria weights, then scales that percentage against the point value you assign. A student who scores 80 percent on a 10-point question receives 8 points for that question.
Step 6: Create the Session
When all settings are configured and your criteria weights total 100 percent, click Create Session. QuizMagic generates a unique session link and QR code.
Distribute the session link to students via your school’s learning management system, email, a messaging platform, or by displaying the QR code on your classroom projector. Students access the quiz on any device without creating an account. They type their essay responses directly into the response field and submit when finished.
Reminder: Anti-cheating monitoring, including copy/paste blocking, activates the moment a student opens the session link, before they begin writing. Students see a brief notification that the session is monitored.
Step 7: Review AI-Generated Scores and Release Results
After students submit, their responses queue in your Smart Sharing results dashboard. Wait until all submissions are in, then start the AI Grading process. You can only run the AI Grading if you have tokens available.
1 token = max of 100 responses (students x questions)
1 token can grade a 4-item essay quiz for 25 students
What the AI provides for each submission
For every essay response, you see the following in the dashboard.
Criterion scores. A score for each active criterion, displayed as a percentage of that criterion’s weight. For example, if Depth of Analysis is weighted at 35 percent and the student scores 70 percent on that criterion, they receive 24.5 of the possible 35 points from that criterion.
Composite score. The weighted total across all active criteria, expressed as a raw score against the question’s point value.
Narrative feedback. A written summary explaining the score on each criterion, what the student did well, where their response fell short, and specifically what a stronger response would have included. This is the feedback that goes to the student when you release results.
How to review and finalize
Click any submission to open the full review panel. Read the student’s response alongside the AI’s criterion scores and narrative feedback. Then do one of the following.
Accept the AI evaluation. If the scores and feedback accurately reflect the quality of the response, accept without changes. The composite score stands as submitted.
Adjust individual criterion scores. Use the slider next to each criterion to raise or lower the AI’s suggested score. The composite score recalculates automatically as you adjust.
Add your own comments. Type your personal feedback in the teacher comments field. Your comments appear alongside the AI-generated feedback when the student reviews their results.
Work through all submissions before releasing. The Grade Release Toggle holds all scores and feedback in draft mode until you are ready. When you have finalized every submission, toggle the release to publish results to all students simultaneously. If a student has a QuizMagic account, even a free account, the result will appear in the student’s Student Dashboard section.
Time estimate: Reviewing and finalizing an AI-generated essay evaluation typically takes 30 to 90 seconds per submission, compared to 8 to 15 minutes for manual grading from scratch. For a class of 30 students, that compresses three to four hours of grading into 20 to 45 minutes of review.
Tips for Accurate AI Grading
Write specific essay questions/prompts. The AI evaluates responses against the question/prompt you wrote. A vague prompt produces responses that the AI cannot evaluate precisely. Specific prompts that name the concepts, relationships, or arguments you want students to address produce more accurate criterion-by-criterion scores.
Match criteria to the assignment purpose. A grammar-heavy weighting on a content-knowledge assessment sends the wrong signal to students and produces feedback that is not useful for teaching the content. Weigh the criteria that actually reflect the learning objective for this assignment.
Use Factual Accuracy only when the correct facts are in the source material. If you generated the essay quiz from a specific PDF or set of slides, the AI has the source content as a reference for evaluating factual claims. If the quiz was generated from a broad topic prompt, Factual Accuracy scores depend on the AI’s general knowledge rather than your specific curriculum, which may introduce inaccuracies in the evaluation.
Review the first batch carefully. The first time you use AI essay grading for a new assignment type, spend more time reviewing each submission than you normally would. This calibration pass helps you understand where the AI’s evaluation aligns with your judgment and where it consistently diverges, so you can adjust weights for future sessions.
Combine essay grading with analytics. After releasing results, review the per-criterion averages across the class in your analytics dashboard. If 70 percent of students scored poorly on Depth of Analysis, that is a teaching signal, not a student failure. Use the class-level criterion data to plan your next lesson. See How to Read Your Quiz Analytics Dashboard for a full walkthrough.
Troubleshooting Common Issues
The Grading Criteria panel does not appear in Smart Sharing
The Grading Criteria panel only appears for essay quiz types. If you are sharing an objective quiz (MCQ, True/False, Fill-in-the-Blanks, Short Answer), the panel is not shown because those question types use direct answer matching rather than AI essay grading. Confirm that your saved quiz uses the Essay question type.
The weights do not add up to 100 percent
The panel shows a running total below the criteria list. The Create Session button stays inactive until your weights total exactly 100. Check for a criterion that is checked but has a value of 0. It counts as active with zero weight, which may throw off your total. Either uncheck it or assign it a weight.
AI grading results seem inconsistent with the quality of the response
This usually happens for one of three reasons. First, the prompt was too vague for the AI to evaluate the response precisely. Rewrite the prompt for future sessions to be more specific about the concepts students should address. Second, the criteria weights do not match the actual learning objective. For example, Coherence is weighted heavily, but the assignment is primarily a content knowledge check. Third, the student’s response was very short or off-topic, which causes the AI to score most criteria low uniformly. These cases are the most valuable ones to review manually and use as calibration examples.
A student’s submission does not appear in the dashboard
Confirm that the student accessed the quiz via the Smart Sharing session link and completed and submitted the quiz rather than closing the window mid-response. Partially completed submissions that were not formally submitted do not appear in the results dashboard.
Frequently Asked Questions
Is AI essay grading available on the Free plan or the premium trial? No. AI essay grading requires an active Premium subscription and uses a separate token system. Essay grading tokens are not included in the free premium trial. This is the one Premium feature that the trial does not cover. Upgrade at quizmagic.io/pricing to access essay grading.
How many essays can I grade per month on the included tokens? Each Premium subscription includes 5 tokens per month, and each token covers up to 100 essay responses. This means the included allocation covers up to 500 essay submissions per month at no additional cost.
What happens if I run out of tokens mid-month? When your tokens are depleted, the AI grading engine is unavailable for new sessions until you purchase additional tokens or until your tokens reset at the start of your next billing cycle. Additional tokens cost $1.00 each and are available in your account settings. Each additional token covers 100 more essay responses.
Do unused tokens roll over to the next month? No. Monthly token allocations reset on your billing date, regardless of how many tokens were used. Purchased additional tokens, however, do not expire and remain available until used.
Can I mix essay questions with other question types, like MCQ, in the same quiz? Not currently. Essay quizzes are a separate quiz type because the grading approach is fundamentally different from objective question matching. MCQ, True/False, Fill-in-the-Blanks, and Short Answer questions can be combined in a Mixed Mode quiz, but Essay questions currently require their own dedicated quiz session.
Do students see the grading criteria weights before submitting? No. Students see only the essay questions/prompt and the response field. The grading criteria and their weights are visible only in your Smart Sharing configuration panel and in the results dashboard after submission.
Can I change the grading criteria after the session has started? No. Grading criteria are locked when you click Create Session because the AI evaluates every submission against the same criteria for consistency. To use different criteria weights, create a new Smart Sharing session for the same quiz with the updated configuration.
