In Short:

• Upload a single class PDF and EssayHero marks every essay in parallel, with results in minutes
• You always have the final say: override any score, add notes, approve or send back for review
• Build custom rubrics, store exam questions, and reuse comment templates across sessions
• Class analytics show you score distributions, per-criterion breakdowns, and a student heatmap
• Everything exports to CSV, print, or individual share links for students

The Problem Every Teacher Knows

Marking a class set of essays takes hours. Thirty compositions, each needing careful reading, criterion-by-criterion assessment, written feedback, and a score. It's some of the most important work a teacher does, and it's also some of the most time-consuming.

I've been there. Over years of marking essays from Form 4 students through to university dissertations, I know the pattern: you start strong, give detailed feedback on the first ten, then fatigue sets in. The last few essays get shorter comments. The consistency slips. It's not a failure of effort — it's a limitation of time and energy.

EssayHero's batch marking suite is designed to change that equation. Not by replacing your judgement, but by doing the heavy lifting so you can focus on the parts that actually require a teacher: the overrides, the nuanced comments, the patterns you spot across a class that no AI would notice.

How Batch Marking Works

The workflow is a seven-step wizard. It sounds like a lot, but most steps take seconds.

Configure. Choose the exam type (HKDSE Paper 2, IELTS Task 1 or 2, IGCSE, or a custom rubric you've built). Set the feedback verbosity — brief if you just want scores and key points, standard for balanced feedback, or detailed if you want exemplar rewrites for targeted students. Add any custom instructions you want the AI to follow ("focus on paragraph transitions" or "be strict on register").

Upload. Upload one PDF containing all your students' work. This is typically a stack of scanned handwritten essays, though typed work in a single document works too.

Review pages. The OCR engine processes each page and shows you the extracted text alongside the original scan. You can correct any misreads. Word count badges flag pages with suspiciously low word counts, which usually means the OCR struggled with the handwriting.

Review groupings. The AI detects where one essay ends and the next begins, and attempts to extract student names. You'll see proposed groupings that you can adjust — merge pages, split them, rename students. The system flags groups with very low word counts so you can check whether a page was missed.

Analyse. Click go and every essay is sent to Gemini AI in parallel. For a class of thirty, this typically takes two to four minutes. You see progress in real time as each essay completes.

Results. Every essay has a full analysis: scores per criterion, paragraph-by-paragraph feedback, language corrections, and suggestions. From here, the real teacher work begins.

The Teacher Grading Suite

The analysis is a starting point, not a final answer. The grading suite gives you seven tools to make the results yours.

Score override

Click any criterion score to change it. If the AI gave a student 5 for Content but you think it's a 4, change it. Your override is stored separately from the AI score, so you can always see what the AI thought versus what you decided. Overrides flow through to CSV exports and share links — students see your score, not the AI's.

Review workflow

Each essay starts as "pending". As you work through the class, mark essays as "reviewed" (you've looked at it) or "approved" (ready to return to the student). Filter the list by status so you can focus on what still needs attention. Bulk actions let you approve all reviewed essays at once when you're satisfied.

Feedback verbosity

Not every essay needs the same depth of feedback. Brief mode gives you scores and headline comments, using roughly 60% fewer tokens. Standard mode is the balanced default. Detailed mode includes exemplar rewrites showing the student how a paragraph could be improved. You can set this per session, so a quick formative check gets brief feedback while a summative assessment gets the full treatment.

Custom instructions

A free-text field that gets injected directly into the AI prompt. This is where your expertise shapes the feedback. Tell the AI to focus on a particular weakness you've been working on in class, or to be especially strict on a criterion you've been emphasising. The AI doesn't know your teaching context — this is how you give it context.

Custom rubrics

The built-in rubrics cover HKDSE, IELTS, and IGCSE. But if you teach a different curriculum, or if your school uses modified criteria, you can build your own. Fork an existing rubric as a starting point or create one from scratch. Define your criteria, set the score range for each, and save it. Your rubrics persist in the database and appear alongside the built-in options every time you start a new batch.

Class analytics

Once all essays are marked, the analytics panel gives you three views. A score histogram shows the distribution of total scores across the class — you can see at a glance whether the set skewed high, low, or clustered around a particular band. A per-criterion bar chart compares class averages across criteria, making it obvious if the whole class struggled with organisation but did well on content. And a student-by-criteria heatmap lays out every student and every criterion in a colour-coded grid, so outliers and patterns jump out immediately.

All of this is pure CSS — no charting library, no JavaScript overhead. It loads instantly.

Enhanced export

When you're done, there are several ways to get results out. CSV export includes every score (with overrides flagged), teacher notes, review status, and student names. Print mode generates a page-break-separated report for every student, ready to hand back. And you can generate individual share links that give each student a private URL to view their full analysis online.

Question Bank and Comment Templates

Two features that save time across sessions, not just within them.

Question bank

If you teach the same exam type repeatedly, you're assigning from the same pool of past paper questions. The question panel lets you store questions with their full prompt text and text type. Add them manually or, if you have a scanned question paper, use the OCR to extract the text. Questions are saved to your account and loaded automatically when you start a new batch with the same exam type.

When you assign questions to essay groups, the AI receives the question context alongside the student's response. This means the feedback is specific to what the question actually asked, not generic criteria assessment. It's the difference between "your argument lacks evidence" and "you haven't addressed the counterargument that the question explicitly asks you to consider".

Comment templates

Teachers develop stock phrases over time. "Good use of topic sentences throughout." "Register inconsistency — formal and informal language mixed." "Your conclusion introduces new information instead of summarising." The comment bank lets you save these as templates, organised by category: grammar, structure, content, vocabulary, or general.

When you're writing teacher notes for a student, open the comment bank and click to insert. It's faster than typing the same feedback for the fifth time that session. Templates are saved to your account and available across all future sessions.

Session History

Every batch marking session is saved automatically. You can return to past sessions to review results, check which students you've approved, or export data you forgot to download. Sessions store everything: the configuration, the OCR text, the groupings, the analysis results, your overrides, and your notes.

Over time, this builds into a record of your marking across the term. If you mark the same class multiple times, you can look back and see whether scores are trending upward, which criteria are improving, and which students are making progress.

Built for HKDSE, Works for Any Exam

The primary focus is HKDSE Paper 2, because that's the exam most of our Hong Kong teachers are marking. The criteria, level descriptors, and scoring are all built in. But the system is exam-agnostic by design. IELTS Writing Task 1 and Task 2 are fully supported. Cambridge IGCSE is available. And custom rubrics mean you can use it for any assessment framework you teach.

The same workflow applies regardless of exam type: upload, review, analyse, grade, export. The AI adapts its feedback to whichever criteria you've selected.

What It Isn't

I want to be honest about the boundaries.

This is not a system that replaces teacher marking. The AI provides a first pass — a detailed, criteria-aligned first pass — but the scores and feedback are estimates. The review workflow exists precisely because we expect you to disagree with the AI sometimes. When you override a score, that's the system working as intended.

The OCR is good but not perfect. Handwriting varies enormously, and some pages will need manual correction. The boundary detection gets it right most of the time, but you should always check the groupings before analysing.

And the feedback, while structured and specific, lacks the context that comes from knowing a student over time. It can't say "this is a big improvement from last month" or "I know you can do better than this". That's your job.

Getting Started

Sign in to EssayHero and go to /batch. Upload your class PDF and follow the wizard. The whole process from upload to results typically takes under ten minutes for a class of thirty.

It's free. No usage limits on batch marking. If you use it with your students, I'd genuinely like to hear how it goes — what works, what doesn't, what would make it more useful for your classroom.

Questions or feedback? Email hello@essayhero.app.