What Happens When You Ask Students About AI And They Tell You the Truth?

The devices are winning. Grades are covering for it.

Three-quarters of high school students say they misuse technology during class every day. A quarter say they've used AI to help them during a quiz or a test, an activity that’s supposed to measure what a student actually knows.

Those numbers come from a new Thomas B. Fordham Institute survey of just over 2,000 current and recent high schoolers.

The report is worth more than passing attention because the findings aren't coming from watchdog researchers or worried parents. They're coming from students, describing their own classrooms.

The picture the survey builds isn’t really about phones, or even about AI on its own. It's about how those two things interact with school-issued devices and with grading policy to erase whatever used to keep them in check.

Two-thirds of students say YouTube is freely accessible on their school laptop. Nearly half say they've used AI to solve math problems outright, more than a third to summarize a book so they don't have to read it, and a quarter to help during an exam.

Sixty-two percent of students spend the majority of their class time on a school-issued device. The more time students spend on those devices, the more likely they are to believe missing class won't hurt their grade.

Meanwhile, the guardrails that might offset all of this are thin. Twenty-nine percent of students say they get partial credit for assignments they never turned in. Most say they can retake a test at least once if they don't like their score.

All this taken together makes this more than just another AI story. It becomes a story about trust in the K-12 public education system.

And what we see on that issue is that public confidence in K–12 education is at a record low by three separate measures. There’s the Gallup and Walton Family Foundation satisfaction poll, the PDK poll, and Gallup's separate institutional-confidence survey.

It's tempting to write all that off as national mood, disconnected from any particular school. But Fordham's data suggest at least part of it is earned. Fifty-seven percent of surveyed students report earning all or mostly A's, including 44 percent of those who scored under 1000 on the SAT. That’s below the 2025 national SAT average of 1029, meaning that group sits squarely in the bottom half of the national achievement distribution.

A report card that says "doing fine" while a third of the room is on a screen and a quarter had AI help on the test isn't lying, exactly. It's just not measuring anything a family can rely on.

Nor is this a U.S.-only pattern. The recently released 2025 PISA results show reading scores falling in three-quarters of participating countries, with U.S. math and reading down the equivalent of three-quarters of a school year since 2018.

The sharpest declines are concentrated among students who use AI to draft their written work. That's consistent with what AEI's Survey Center found this summer. Forty-two percent of teens use an AI chatbot at least weekly, and homework is the single most common reason.

This isn't an isolated classroom-management problem. It's what a generation's relationship with a new technology looks like when it arrives faster than the policies meant to govern it.

None of that means AI has no place in a classroom, or that devices should disappear. It means the current arrangement, including unrestricted access, weak assessment, and grading that absorbs the consequences, isn't sustainable. Students are the ones telling us so.

Districts have real, specific options available to them. Here are five of them.

—Set and enforce actual limits on device time, rather than leaving it to individual teachers' discretion.

—Shift more weight onto in-class, paper-and-pencil, and oral assessment, where AI assistance isn't available and what a student knows is directly observable.

—End routine test retakes and partial credit for unsubmitted work as district-wide policy, not teacher-by-teacher practice.

—Audit school-device firewalls, since two-thirds of students say YouTube gets through them.

—Pair any phone ban with an actual device policy, since the data show one without the other barely moves the numbers.

None of this requires waiting for a perfect AI-detection tool or a state mandate. It requires districts to treat what students said in this survey as an honest diagnosis rather than background noise.

__________________________

Bruno V. Manno is a senior advisor at the Progressive Policy Institute, where he directs the What Works Lab, and is a former U.S. Assistant Secretary of Education for Policy. You can find him on LinkedIn.