Harvard report: AI suspicion is eroding student-teacher trust

What the report is and who wrote it

An AI Policy Isn't a Playbook: Educators Need Agency, Not Just New Rules was written by Beck Tench, Emily Weinstein and Carrie James, with Allison Starks and Sara Konrath. The report states that it was published by the Center for Digital Thriving in September 2026. The centre is based at Project Zero at the Harvard Graduate School of Education. The work was funded by a grant from the John Templeton Foundation, with further support from the Susan Crown Exchange and Pivotal Ventures.

The evidence comes from three places, and it helps to keep them apart.

The methods section of the report gives completion rates of 51.7% for teachers and 30.2% for principals. The percentages below are from the survey unless stated otherwise.

What the survey found

Teachers and principals were asked whether they had experienced any dilemmas or difficult decisions related to generative AI in their school or classroom. According to the report, 25% of teachers and 29% of principals who answered said yes. The share rose with the age of the students.

The 74% and 69% figures are sometimes reported as shares of all teachers. They are not. They are shares of the written descriptions given by the educators who had said yes to the dilemma question. The report notes that elementary teachers make up about half of US public school teachers, which pulls the overall 25% figure down.

One methodological detail is worth knowing. The report says the 256 written descriptions were coded by one human and two AI models, and that the three agreed on 93% of them.

Detection, proof and students who hold back

The report's sharpest section is about proof. It says tools marketed to detect cheating "have been shown to be unreliable", which leaves teachers who suspect a student with no reliable way to show it. The authors coin the word "AI-dar" for a teacher's hunch that a piece of work was not written by the student.

A seventh grade English teacher quoted in the report put it this way: "When a student digs in and says they didn't use it, there's no actual proof." A ninth grade English teacher described running a suspicious piece through a detector, getting a result that it was written by a human, and still not believing it.

Students described the same problem from the other side. The report says the researchers heard many stories from young people about being falsely accused, and from some who now limit their effort on purpose. One 15-year-old said: "We can't make ourselves sound too smart anymore, we cannot give our full effort anymore, because then we're at the risk of being detected for AI." An 18-year-old college student described putting a group paper through four detectors and finding that one of them flagged all of the student's own writing.

The report calls the result "two-way suspicion", defined as two parties suspecting each other of using AI, which erodes trust whether or not AI was used. A 16-year-old quoted in the report described it as "kind of like a hellscape for both students and teachers".

What the authors said

Beck Tench, a co-founder of the centre, told Education Week that adult debate about AI has centred on students handing their thinking over to a chatbot. She said the report shows something else as well: "there's also this cognitive withholding that some students are doing in order to evade being falsely accused in environments where accusations don't get any kind of conversation." On the back-and-forth between detection tools and workarounds, she said: "It doesn't breed trust. It breeds a race to the bottom."

Speaking to ABC News, Tench said: "Teachers can't see their students anymore." She said teachers find it harder to tell what students know and no longer get the questions students would once have asked them, because those questions now go to chatbots. Lydia Bach, a 12th grade student interviewed by ABC, said: "There's always been this sort of clashing between students and teachers when it comes to trust, but AI has just enabled that to grow."

The report does not recommend a detector, a ban or a model policy. It offers five free classroom activities, which it calls plays, and three changes of approach for teachers: from assuming to asking, from tech-centred to student-centred, and from "us-vs-them" to "us-and-them". The authors' argument is that teachers do not need to become AI experts to talk to students about how and why they use these tools.

What this means for students and teachers

• Students: if you are worried about being flagged, keep your drafts, notes and version history. They are evidence of how a piece was written, and they are more useful than writing below your ability.
• Students: ask each teacher what AI use is allowed on each assignment. The survey suggests most schools had no specific policy in spring 2025, so the answer may differ from class to class.
• Teachers: a detector result and a hunch are both starting points for a conversation. The report's accounts show the harm done when an accusation comes with no discussion.
• Teachers: say in advance what you will do if you suspect AI use, including whether the student will get a chance to explain or to show their process.
• School leaders: fewer than a quarter of principals reported a specific generative AI policy. Teachers quoted in the report describe being left to decide alone.

FAQ

What happens next

The report's survey data is now more than a year old, and the authors describe their youth interview project as ongoing. The useful part for schools this term is the description of the mechanism: suspicion without proof on one side, and no way to prove innocence on the other. We will update this article if the centre publishes newer survey figures.