The short answer
- AI made fluent prose free, so a rubric that rewards grammar, structure, and surface polish now rewards a prompt — not a student.
- Shift weight to what AI can’t do for them: original analysis, use of specific class material, personal reasoning, and process.
- Grade the process, not only the product — outline, drafts, and reflection reveal thinking that a final essay can hide.
- Reward specificity and stance — the particular example, the defended opinion, the connection to your class are hard to outsource.
- This isn’t lowering standards; it’s aiming them at the skills that were always the point.
Here’s an uncomfortable thing to notice about your current rubric: if a large share of the marks go to grammar, organization, and fluent expression, then a student with a good prompt can now score well without thinking at all. AI made those surface qualities free. A rubric that still rewards them heavily isn’t measuring the student anymore — it’s measuring the tool. The fix isn’t to lower standards; it’s to point them at what was always the real target.
What did AI actually change?
It commoditized the surface and left the substance untouched. Grammar, structure, transitions, a competent five-paragraph shape — these used to be reliable proxies for effort and skill, because producing them took work. AI removed the work. What it did not remove is the ability to think: to analyze something original, to marshal specific evidence, to take and defend a position, to connect ideas to a particular class and context. Your rubric should follow the value — away from the proxy that’s now free, toward the substance that isn’t.
Where should the marks move?
Toward four things AI can’t do for a student:
| Reward more | Reward less (as the main event) |
|---|---|
| Original analysis and insight | Fluent, grammatical prose |
| Specific evidence — from your class, texts, data | Generic, correct-sounding generalities |
| A defended point of view | A balanced summary with no stance |
| Evidence of process (drafts, reflection) | Only the polished final product |
None of this means grammar stops mattering — it’s a baseline for communication. It means grammar stops being where the points are. A submission that’s beautifully written and says nothing specific should no longer earn a strong grade, because “beautifully written” is now the cheap part.
How do you grade thinking directly?
Assess the process and demand the specific. Concretely:
- Weight process artifacts. Give real marks to the outline, the draft, and a short reflection. Thinking is visible in the mess; polish hides it.
- Require class-specific connection. “Draw on Tuesday’s discussion of X.” A model trained on the open internet can’t, and a student who was there can.
- Reward the particular. The specific example, the named counterargument, the personal reasoning — score these explicitly. Generic is now suspicious; specific is now valuable.
- Use oral check-ins where you can. Two minutes of “walk me through your argument” grades understanding directly. Expensive in time, but unfakeable.
Is this fair to students?
It’s fairer than the old rubric — it just moved the goalposts back to where they belonged. The student who thinks carefully but writes plainly was always undersold by a fluency-weighted rubric; the AI era finally makes that visible. Rewarding analysis, specificity, and process advantages the students actually doing the intellectual work, and disadvantages the ones outsourcing it — which is the point of grading in the first place.
Make the new criteria explicit and exemplified so they’re no more subjective than the old ones, and you’ve built a rubric that measures the student instead of the tool.
Pair this with AI-resistant assignment design and a clear AI policy — together they form a system where the tasks, the rules, and the grading all point at the same thing: the thinking only the student can do.
Frequently asked questions
How should rubrics change because of AI?
Shift weight away from surface features AI now produces for free — grammar, structure, fluency — and toward the thinking it can’t do for a student: original analysis, use of specific class material, personal reasoning, evidence of process, and a defended point of view. The skills a rubric rewards should be the ones a prompt can’t fake.
Should I stop grading grammar and structure?
Not entirely — they still matter for communication. But they should no longer be where the marks are concentrated, because AI can supply them on demand. Keep them as a baseline expectation and move the bulk of the grade to analysis, specificity, and process, so a fluent but empty submission can’t score well.
How do I grade thinking instead of writing?
Ask for and assess process artifacts (outlines, drafts, reflections), require connections to specific class discussions or texts, reward original examples and defended opinions, and use oral check-ins where practical. These target reasoning directly rather than the polished prose that reasoning used to be visible through.
Won't this make grading more subjective?
It can, so make the new criteria explicit: define what strong analysis, specificity, and process look like at each level, with examples. Well-described criteria for thinking are no more subjective than well-described criteria for writing — and they measure what you actually care about instead of what a chatbot can supply.