A teacher who uses an AI essay grader should know which decisions stay with the teacher. EssayGrader (essaygrader.ai), a rubric-based AI essay grading platform for K-12 and higher-education teachers, states its answer in two public places, the AI Grading Accuracy Center and the Trust Center. This article reports that wording and turns it into a short list of what the grader drafts and what the teacher decides.
The Accuracy Center says the teacher approves everything. A student sees no grade and no comment until a teacher releases it. The Trust Center describes the same rule from the data side: AI-generated grades and feedback arrive as a draft, nothing is recorded or shared with a student until the teacher reviews it, and the teacher can accept, edit or discard any part. The Trust Center adds that there is no fully automated decision-making in the product.
| Decision | EssayGrader drafts | The teacher decides |
|---|---|---|
| The criteria | Nothing. The rubric comes from the teacher or the 800+ library. | Which rubric applies and what each level means |
| The score | A score for each criterion, with the reasoning | Whether to keep, change or discard it |
| The comments | Rubric feedback in the Hattie and Timperley style | Which comments a student reads |
| Release | Nothing | When anything reaches a student |
EssayGrader includes AI-writing detection, and the Accuracy Center states plainly that it is a signal, not a verdict. It goes further: no AI detector on the market is accurate enough to justify an academic integrity decision on its own. The Accuracy Center adds that paraphrased AI text defeats every detector it has tested, EssayGrader's own included.
Detection reports go to the teacher, never automatically to the student or a parent, and they show evidence rather than a verdict. The teacher reads the highlighted passages, looks at the writer's earlier work and decides whether a conversation is needed.
EssayGrader says it is not an official scoring engine, and its scores do not replace the state's own. The Accuracy Center's 94% criterion-level agreement figure comes from a July 2026 study of 22,000+ comparisons, with the AI score within one point of the official score. The same page says the benchmark sets are cleaner than a typical classroom stack, so 94% is an upper bound on alignment, not a class average.
The Accuracy Center lists the cases where a teacher should look hardest.
| Case | What the Accuracy Center says |
|---|---|
| Very short essays | Under roughly 75 words there is not enough evidence for reliable criterion-level scoring, so these essays are flagged rather than scored with false confidence. |
| Creative and highly personal writing | It can score lower than it deserves. |
| Factual claims outside the source material | EssayGrader cannot say whether such a claim is true. |
| Handwriting | Handwriting reduces accuracy for any AI grader, so the transcription sits beside the scan for the teacher to check. |
Let the grader do the first pass on student writing,. Then read the scores that matter: the borderline ones, the very short ones, the personal ones and anything with a detection flag. Keep the final grade, the final comment and any integrity decision in your own hands.
The setup fits how teachers already work. A grade a teacher has read and released belongs to that teacher, whichever tool produced the first draft.