For decades, grammar, mechanics, and organization served as reasonable stand-ins for writing ability, and many rubrics and scoring systems rely heavily on them. Research from a major testing organization has shown how that assumption breaks down: AI-generated essays outperformed human-written essays on language-related features, and an automated scoring system gave the AI essays higher scores than human raters did. The lesson for classroom teachers is direct. If a rubric rewards fluency heavily, it will reward students who let a tool supply the fluency.

This does not mean conventions no longer matter, since clear writing remains an important goal. It means the relative weight of conventions in a grade should reflect their value as evidence of a student's own skill. When any student can produce clean sentences instantly, the more informative signals are the quality of the claim, the choice and use of evidence, and the logic connecting them.

Many teachers already sense this and are quietly adjusting how they grade. Making the adjustment explicit in the rubric helps in three ways. It tells students what you value, it makes your grading more consistent, and it keeps your feedback focused on the aspects of writing that you can actually teach. A tenth-grade teacher, for example, might raise the weight of evidence and reasoning from twenty to forty percent while lowering conventions to ten.

Rebalance the point distribution

Look at your current rubric and add up the share of points assigned to reasoning criteria versus surface criteria. If sentence fluency, grammar, and formatting together carry thirty or forty percent of the grade, consider shifting some of that weight toward claim, evidence, and analysis. A common pattern is to give reasoning criteria around sixty to seventy percent and reserve the remainder for organization and conventions.

• Claim: takes a specific, arguable position that responds to the prompt.
• Evidence: selects precise, relevant details or quotations and explains why they were chosen.
• Reasoning: explains how the evidence supports the claim and addresses complexity or counterargument.
• Organization: sequences ideas so each paragraph builds the argument.
• Conventions: keeps errors from interfering with meaning, weighted lightly and graded after the content.

Write descriptors that reward specifics

Polished but empty writing tends to share certain features: general claims, vague references to the text, and conclusions that restate the introduction. Descriptors can target those features directly, such as at the highest level, the response cites at least two specific details from the source and explains what each reveals about the claim. A student who relies on fluent generalities cannot meet that standard, and a student who understands the material can.

Anchor papers help students see the difference. Show a smooth but shallow paragraph beside a rougher but insightful one, and ask students which earns the higher score and why. The comparison teaches that the rubric rewards substance, and it often prompts students to take intellectual risks they might otherwise avoid. Rotating the anchors each term keeps the exercise fresh and guards against students simply copying a familiar model.

Make sure your grading tools follow the same priorities

Any scoring tool, human or automated, should apply the weights you set instead of defaulting to surface features. If you use an AI-assisted first pass, confirm that it scores each criterion against your descriptors and that it can explain its reasoning for the evidence and analysis scores. Spot check a few essays where polish and substance diverge, since those are the cases most likely to reveal a mismatch.

A rubric-based approach that lets teachers define their own criteria is well suited to this kind of rebalancing, because the teacher controls the weights and descriptors. The tool applies them consistently across a large set, while the teacher reviews edge cases and adjusts scores. This keeps the final judgment where it belongs. A teacher who notices the tool rewarding fluent but empty prose can simply revise the descriptor and rerun the set the same afternoon.

Support students who are still developing fluency

A reasoning-first rubric can be fairer to students who think well but write with errors, including multilingual learners and students with writing-related disabilities. Their ideas receive credit even when surface features are rough. Feedback can then address conventions separately, with targeted suggestions that do not overshadow the strengths in their thinking. Teachers who adopt this approach often find that grade distributions become more accurate reflections of what students actually understand.

Communicate the change to students and families so no one is surprised. A short explanation, such as this year the rubric gives more weight to evidence and reasoning because those are the skills that matter most in college and work, is usually well received. Over time, students learn that polish without substance does not earn credit. Sharing the new rubric before the assignment is assigned, along with one scored example, removes most of the anxiety and the later grade disputes.