A new research report on generative AI in schools describes a pattern it calls two-way suspicion, where teachers wonder whether a polished essay is really the student's own work while students worry that their honest writing will be flagged as machine-made. The damage does not depend on whether anyone actually cheated. Once a classroom runs on doubt, comments become defensive, students hide how they use tools, and conversations about writing turn into conversations about enforcement. For English teachers who already read dozens of essays a week, that shift is exhausting and it undermines the relationship that makes feedback land.

Survey data from the past year suggests the worry is widespread on the student side, with roughly half of students in one large national panel saying they feared being falsely accused of cheating with AI. That fear changes behavior in predictable ways. Students stop drafting in shared documents, delete earlier versions, or avoid asking for help because any visible assistance might look suspicious. The result is that teachers lose exactly the evidence of process they would need to judge authorship fairly, which makes the next round of suspicion even harder to resolve.

Trust rebuilds fastest when expectations are specific rather than moralizing. A syllabus line that says AI is not allowed leaves students guessing about brainstorming, grammar checking, or asking a chatbot to explain a prompt. A short chart that lists each assignment type and what is permitted, such as outlining help allowed but sentence generation not allowed, removes the guesswork and lets students make honest choices. It also gives you a clear standard to point to if a conversation about a particular submission ever becomes necessary.

Make the writing process visible

Short, low-stakes checkpoints do more to establish authorship than any detector. A thesis statement written in class on Monday, a planning page turned in Wednesday, and a first draft submitted with its revision history on Friday create a trail that is hard to fake and easy to review. Teachers do not need to grade every checkpoint, since a quick completion mark is enough to make the trail exist. When a final essay suddenly departs from that trail in voice or sophistication, you have specific, nonaccusatory evidence to start a conversation.

• Publish an assignment-by-assignment AI permissions chart on the first day and repeat it on each prompt sheet.
• Collect one in-class writing sample early in the term so you know each student's baseline voice.
• Ask students to submit draft history or a brief process note with every major essay.
• Hold a short conversation, not a disciplinary meeting, when a submission looks out of character.
• Return feedback tied to named rubric criteria so students see that grades come from the writing itself.

Replace guesswork with rubric evidence

A rubric changes the emotional temperature of grading because it points to observable features of the text rather than a teacher's impression. If a paragraph is marked down for lacking specific textual evidence, the student can see exactly what is missing and fix it, regardless of how the draft was produced. This keeps feedback focused on the quality of reasoning and organization, which are the things you can actually teach. It also protects honest students from the feeling that a grade reflects a hunch about their integrity.

Consistency matters here as much as clarity. When the same rubric is applied the same way to the first essay and the hundred and first, students can trust that their grade was not shaped by fatigue or by who they are. Rubric-based AI tools that produce a first-pass score and draft comments against the teacher's own criteria can help hold that line across a large stack, provided the teacher reviews and adjusts every result before it reaches a student. That human review is what keeps the process accountable.

Talk about AI openly in class

Researchers who studied the trust problem recommend open conversation between educators and students, and that advice holds up in practice. Ten minutes spent asking students how they have used AI for school, what felt helpful, and what felt like a shortcut tends to surface more honest information than any policy memo. Students often describe uses teachers had not considered, such as asking a chatbot to quiz them on a reading or to explain a confusing rubric term. Those answers help you refine your permissions chart so it matches how students actually work.

It also helps to be candid about your own use of technology. If you use AI to draft first-pass comments that you then edit, saying so models the same transparency you expect from students. Students tend to respond to that candor by giving the same in return, especially when you explain that a person always reviews the final feedback. The goal is a classroom where tools are discussed as tools, and where nobody has to hide anything to feel safe.

What to do when suspicion surfaces anyway

Even with good systems, a submission will occasionally raise real questions. Start with the evidence you already have: the process trail, the in-class baseline, and the rubric criteria the essay does or does not meet. Invite the student to talk through their argument, explain a word choice, or revise a paragraph while you watch, because genuine authorship usually shows quickly in conversation. Avoid treating a detector score as proof, since those tools produce false positives that can seriously harm a student's record.

Document what you did and why, in neutral language, so that any follow-up by a colleague or administrator can see your reasoning. Over time, a consistent and fair response teaches students that the classroom is a place where honest effort is recognized. That reputation is worth more than any single case you win or lose. Trust built this way also makes everyday feedback more effective, since students who feel believed are more likely to act on your comments.