Marking and Feedback at Scale

Here's the core approach. It's simple, and it works for essays, short answers, practical write-ups, and extended responses across any subject:

Step 1: Feed in your rubric. Paste or type your marking criteria, success criteria, or mark scheme. Be specific — include grade boundaries, key indicators, and what you're looking for at each level.

Step 2: Feed in the student response. Paste the student's work (or a typed-up version of it). One piece at a time, or batch several together.

Step 3: Ask for draft feedback. AI generates feedback aligned to your rubric — strengths, areas for development, and specific next steps.

Step 4: Review, adjust, and personalise. This is non-negotiable. You read the feedback, tweak it to match what you know about that student, and make it yours.

The key word in this entire workflow is draft. AI produces a draft. You produce the final version. The AI saves you from staring at a blank comment box for the 30th time — but your professional knowledge of that student, their context, and their journey is what makes the feedback meaningful.

The way you prompt AI should change depending on what the marking is for.

Formative feedback (ongoing, developmental): You want comments that help students improve right now. Focus on specific, actionable next steps. Use prompts like: "Identify two strengths and one specific area for improvement. Give the student a concrete action they can take in their next piece of work."

Summative feedback (end of unit, formal assessment): You want comments aligned tightly to grade descriptors and assessment criteria. Use prompts like: "Assess this response against the following mark scheme. Award a mark and justify the grade with reference to specific criteria. Note what the student would need to do to reach the next grade boundary."

Peer assessment preparation: You can use AI to generate model answers at different levels, so students can see what a Grade 4, Grade 6, and Grade 8 response looks like before they assess each other's work.

Getting the prompt right for the purpose saves you from having to rewrite the AI's output. Be explicit about what kind of feedback you need.

Here are ready-to-use prompt templates. Copy them, adjust for your subject, and start saving hours.

Essay feedback (formative):

"Here is my marking rubric: [paste rubric]. Here is a student's essay: [paste essay]. Provide feedback with: two specific strengths with evidence from the text, one area for development with a specific example from their writing, and one actionable target for their next essay. Write in second person ('you'). Tone: warm but honest. Keep it under 150 words."

Short answer marking (summative):

"Here is the mark scheme: [paste mark scheme]. Here are 10 student responses to this question: [paste responses]. For each student, award a mark out of [X] and write a one-sentence justification referencing the mark scheme."

Peer assessment preparation:

"Generate three model answers to this question: [paste question]. Write one at a low level, one at a mid level, and one at a high level. Use the following success criteria to differentiate: [paste criteria]. Label each clearly. These will be used by students to calibrate their peer assessment."

Here's something that matters more than most teachers realise: students know when feedback doesn't sound like you. If you normally write "Brilliant use of evidence here, Aisha — this is exactly what I was looking for" and suddenly your comments read like a corporate report, they'll notice. And they'll care less.

Two ways to keep your voice:

Option 1: Give AI your voice. Paste in 3-4 examples of feedback comments you've written before. Tell AI: "Match this tone and style when generating feedback. Notice how I address students, the level of formality, and how I phrase praise and targets." AI picks up on your patterns quickly.

Option 2: Use AI for content, add voice yourself. Let AI identify the strengths, weaknesses, and targets. Then rewrite the opening and closing in your own words. This is faster than writing from scratch but still sounds unmistakably like you.

Either way, the goal is the same: feedback that feels personal, because it is. AI handles the heavy lifting of analysis. You handle the human connection.

That conversation produced three pieces of personalised, rubric-aligned feedback in about 30 seconds. Feedback that references the student's actual words, identifies genuine strengths, and gives specific, actionable targets.

Doing that manually would take 15-20 minutes. Across a full class of 30, that's hours saved.

But — and this is critical — you must review every piece of feedback before a student sees it. Here's why:

AI can misread student intent. It can miss cultural context. It can praise something you wouldn't, or overlook something important. It doesn't know that Jake has been struggling with confidence this term, or that Fatima is applying for a scholarship and needs to be pushed harder.

The rule is simple: AI drafts, you decide. Every comment gets your eyes before it gets to a student. No exceptions. This isn't optional — it's a professional and ethical responsibility.

You don't have to overhaul your entire marking process today. Start with one class. One set of books. One assignment.

Try it with formative comments first — the stakes are lower and you'll get comfortable with the workflow before you use it for summative assessments.

Keep a copy of your rubric saved somewhere you can paste it quickly. Build a small library of prompt templates for your most common assessment types. Within a week, you'll have a marking workflow that saves you hours — and your students will still get the thoughtful, personalised feedback they deserve.

Because here's the truth: the best feedback in the world is useless if it sits in your head because you ran out of time to write it. AI makes sure it actually reaches the student.