How do I grade a whole class set of essays?

What should you do before you score the first essay?

The work that makes a class set go quickly happens before you write a single grade. Re-read the task sheet you handed out and the rubric students were given, and decide what you are actually assessing this time. If the task was argument writing, the thesis, the use of evidence and the structure carry the score. Surface errors get a code in the margin, not a paragraph of explanation.

Then read about ten essays without scoring any of them. Sort them into three rough piles as you go: clearly strong, middle, clearly weak. Pick one paper from each pile as an anchor and write yourself a single line about why it sits there. Those anchors become your standard for the rest of the stack. At essay 22, when a response feels borderline, you compare it against a real paper instead of against a memory of what you thought a B looked like an hour ago.

Set a time budget before you start, not halfway through. Decide roughly how long each essay gets and how many sittings the set will take, then put those sittings in your week. A set graded in three planned blocks stays more consistent than the same set graded in whatever gaps appear.

• The task sheet and the rubric exactly as students received them
• Three anchor papers, one per performance band, with a one-line justification each
• A decision about what you are not grading this round
• A per-essay time budget and a plan for how many sittings the set takes
• A short list of the feedback phrases you already know you will repeat

Why is it faster to grade criterion by criterion?

Grading one essay completely, then the next, forces you to hold the whole rubric in your head for every student and to re-enter each criterion from cold. Grading criterion by criterion does the opposite. You read the essay once for sense, then judge one thing at a time across the class: all the thesis statements, then all the evidence handling, then all the structure. The judgment stays narrow, and because you are comparing like with like, the levels you award line up across students.

This also limits the halo effect. A confident opening paragraph makes a reader generous about everything that follows, and a scrappy first page makes a reader harsh. Separating criteria keeps a strong introduction from quietly lifting the evidence score. For short essays you can genuinely do one pass per criterion through the whole stack. For longer pieces, read each essay once but score the criteria in a fixed order with the rubric open in front of you.

Score first, comment second. Deciding a level and composing prose at the same time is the slow part of grading, because you are doing an assessment task and a writing task in the same breath. Record all the scores, then go back and write comments knowing where each essay landed. Most teachers find the comments get shorter and sharper once the grade is already settled.

How do you stop your standard from drifting down the stack?

Drift is predictable rather than shameful. The essays you grade at 10pm on the fourth night get read differently from the ones you graded fresh on Saturday morning, and the papers that sit at the bottom of the pile are usually the same students’ every time. Two habits handle most of it. Shuffle the set before each sitting so the running order changes, and cover names where the format lets you, revealing them when you write the personal note at the end.

Work in blocks of eight to ten essays with a real break between them, and re-read one anchor paper at the start of each block. It takes ninety seconds and it resets the standard you are about to apply. Without that reset, the standard tends to tighten as fatigue builds.

At the end of the set, back-read. Re-score your first five essays against the standard you finished with, without looking at what you gave them. If one moves a level, that is normal noise. If two or three move, your standard shifted and the early block needs a proper second pass. It is the check you would ask a colleague for in moderation, and it takes ten minutes.

• Shuffle the pile between sittings so the same students are not always last
• Hide names for the scoring pass where the format allows it
• Re-read one anchor paper at the start of every block
• Back-read the first five papers once the set is finished
• Flag borderline papers for a second read rather than agonizing over them alone

How much feedback should each essay actually get?

Cap it deliberately. Two things the student did well, named with a quotation from their own writing, and one next step written as an instruction they can carry into the next piece. Long feedback on a returned essay mostly functions as a justification of the grade, and it arrives when a student is least able to use it. Short feedback that points forward gets acted on more often.

Keep the phrasing you repeat in a comment bank, whether that is a document, a set of margin codes, or a feature in whatever platform you use. The value is not that it saves typing, though it does. It is that the wording you use for a weak thesis stays the same for every student who has a weak thesis, so the class hears one consistent message. The sentence you personalize is the one that names what this particular student wrote.

Then look at the set as a whole before you hand anything back. If 15 of 28 essays dropped quotations in without integrating them, that is a teaching problem, not 15 comment-writing problems. Re-teach it in ten minutes of class time, point students to study guides or model paragraphs, and give them a piece of writing to fix it in.

How do you make the grades defensible to students and parents?

Every score should be traceable to two things: the wording in the rubric, and a line in the student’s essay. If you can say that the response sits at the middle band for evidence because the quotations are accurate but unexplained, and you can point at the paragraph where that happens, the conversation is short. If the only justification is your overall impression, the conversation is long and you will probably concede.

Decide a tie-break rule in advance for essays that sit between two bands, and apply it to everyone. Some departments weight the criterion the task was designed to teach. Others hold the anchor papers side by side and ask which one the response resembles more. Either is fine. Deciding case by case is what causes trouble, because the rule quietly shifts with who wrote the essay and how tired you are.

Keep the anchor papers, with names removed and permission where your school requires it. Next year they save you the sorting step, they give a new colleague a calibrated starting point, and they turn department moderation into a comparison of real work rather than a discussion of adjectives.

Where do AI grading tools fit into this workflow?

This is the part where software helps, and it is worth being precise about which part. Most essay grading tools promise time savings, but the useful ones do not try to replace the judgment. They carry the repetitive passes: reading each response against criteria you defined, drafting a comment tied to a specific line, and proposing a level. What they cannot do is know what your class was taught last month, which student is one honest conversation away from giving up, or where your department has agreed to draw the line between two bands.

Jeddle works on that division of labor. You set up your rubric, success criteria and comment banks once, bring in the class set as a batch, and JeddAI drafts criterion-aligned comments and a suggested grade for each essay. Nothing reaches students until you review it and release it, and you can rewrite any comment or change any grade first. In practice you are still reviewing every response, but you are editing a draft rather than composing from a blank page, which is a lighter task per essay.

The same checks apply to any tool of this kind. Does it grade against your rubric or a generic one? Does it show you the evidence in the student’s writing for the level it proposed? Is the draft private until you sign it off, and can you override the grade? Run your own drift check anyway: pick three essays, grade them yourself from scratch, and compare. If you and the tool disagree systematically, the setup needs fixing before the class set does.

• Set the rubric and success criteria yourself rather than accepting a default
• Check that each drafted comment points to real text in the student's essay
• Keep drafts private until you have reviewed and released them
• Spot-check three essays by hand against the drafted grades
• Treat the suggested grade as a starting point you have to be able to defend