Examplary can automatically provide student feedback based on the answers given in assessments, either taken online or scanned in from paper.
The quality of that feedback depends heavily on the instructions you give. Clear, specific guidance tells the AI what good feedback looks like for your subject and your students, and prevents it from falling back on generic language that doesn't help anyone learn. The tips below cover what to include in your feedback instructions to get the most useful suggestions.
Suggest a structure
A consistent structure makes feedback easier for students to read and act on. Tell the AI exactly how you want each piece of feedback to be organised, and it will follow that pattern across every answer.
A structure that works well in most subjects is: 1) what was wrong, 2) why it was wrong, and 3) how to improve. You can adapt this to your own teaching: some teachers prefer to start with what the student did well, others include a worked example at the end. Whatever pattern you choose, be explicit about it in the instructions.
Structure each piece of feedback in three short paragraphs. The first paragraph identifies what was incorrect in the student's answer. The second paragraph explains why it was incorrect, referencing the relevant concept or rule. The third paragraph gives one concrete suggestion for how the student can improve next time.
Specify the tone
The same piece of feedback can land very differently depending on how it's phrased. Tell the AI what tone you want it to use, especially with younger students or sensitive topics where wording matters.
Most teachers ask for feedback that is constructive and encouraging, without being overly negative or dismissive. You might also specify a level of formality, whether to address the student directly, or whether to use first or second person. The clearer you are about tone, the more consistent the feedback will feel across the class.
Write the feedback in a warm, encouraging tone suitable for students aged 12 to 14. Address the student directly using "you". Acknowledge effort where appropriate, but always be honest about mistakes. Avoid sarcasm, exclamation marks, and language that might feel patronising.
Focus on actionable advice
Feedback is most useful when a student can do something with it. Ask the AI to give specific, concrete suggestions for how the student can improve their answer, rather than abstract observations about what went wrong.
For example, "review the order of operations in step two" is far more useful than "your working contains an error". Make this expectation explicit in your instructions, and the AI will lean towards practical next steps the student can actually act on.
Every piece of feedback must end with a specific, actionable next step the student can take to improve. Refer to the exact part of their answer that needs work, and suggest a concrete action — for example, "rewrite the introduction to state your thesis in a single sentence" or "redo step three using the quadratic formula".
Avoid vague or generic feedback
Comments like "good job" or "needs improvement" don't tell a student anything they can use. Instruct the AI to always explain its reasoning: why an answer was right or wrong, and what specifically led to that judgment.
It also helps to tell the AI what to avoid. Phrases like "don't just praise the answer without explaining why it works" or "don't give one-line feedback" steer it away from the patterns that tend to creep in when instructions are too open-ended.
Do not use generic phrases like "good job", "well done", or "needs improvement" on their own. Every comment must explain the reasoning behind it, referencing something specific in the student's answer. Feedback should be at least two sentences long, even when the answer is mostly correct.
Tailor to the subject
Different subjects call for different kinds of feedback. The instructions you write for a maths assessment should look quite different from those for an essay, a language exercise, or a science practical.
For a maths question, you might ask the AI to focus on the problem-solving steps and the calculations, pointing out where the working went wrong and how to set it out more clearly. For an essay question, you'd ask it to focus on the argument structure, the evidence used, and the clarity of the writing. The more specific you are about what matters in your subject, the more relevant the feedback will be.
Focus your feedback on the student's use of evidence, the strength of their argument, and whether they address the question directly. Reward answers that weigh multiple causes against each other rather than listing them. Highlight any factual errors about dates, people, or events.