Overview

Student use of artificial intelligence for homework rose sharply in 2025 — even as more students worried the technology was harming their ability to think critically. That is the headline from a March 2026 RAND press release based on nationally representative survey data from RAND's American Youth Panel.

The survey, conducted in December 2025 with more than 1,000 students ages 12 to 29 enrolled during the 2025–26 school year, captures a moment when AI moved from novelty to daily school infrastructure.

The tension in one sentence

Students are using AI more, trusting it more, and simultaneously reporting that it may be weakening the thinking skills schools are supposed to build.

Key numbers from RAND

  • 62% of middle school, high school, and college students used AI for homework in December 2025, up from 48% in May 2025.
  • 71% used at least one AI tool for any school-related activity.
  • 67% said AI harmed critical thinking for schoolwork, up from 54% earlier in the year.
  • 78% of students who do not use AI said it harms critical thinking, compared with 60% among AI users.

The increase was driven mainly by middle and high school students. College student usage stayed relatively steady.

How students use AI

Chatbots dominated the tool list:

  • ChatGPT — 53% of students
  • Google Gemini — 28%, more than doubling between May and December 2025
  • Writing helpers like Grammarly — 21%
  • Homework platforms like Chegg or Brainly — 15%

The most common tasks were getting better explanations (38%), brainstorming ideas (35%), looking up facts (33%), and drafting or revising writing (33%). Older students were more likely to use AI for these tasks than younger ones.

The critical thinking concern

RAND's findings align with academic research on cognitive offloading. A January 2026 study of 299 STEM students across five North American universities, published on arXiv, found that students who trusted and routinely used generative AI reported significantly lower cognitive engagement — including reflection, need for understanding, and critical thinking.

The OECD's Digital Education Outlook 2026 adds another layer: when students use general-purpose chatbots without pedagogical guidance, they may produce better-looking work without building lasting knowledge. The advantage often disappears on exams when AI access is removed.

A better study workflow

The research points toward intentional use, not abstinence:

  1. Ask for explanations, not answers. Use AI to clarify a concept you are stuck on, then close the chat and explain it yourself.
  2. Generate questions, not essays. Turn lecture PDFs or YouTube videos into flashcards and quizzes — then test yourself without help.
  3. Apply the Feynman technique. Summarize a topic in plain language, identify gaps, and revisit only the weak spots.
  4. Listen and recall. Convert notes to audio with text-to-speech, then pause and summarize what you heard.

That is the workflow Feynman AI: Study & Memorize is built around: source material in, active recall out.

FAQ

Is using AI for homework cheating?

It depends on your school's policy and how you use the tool. Explaining concepts and generating practice questions is different from submitting AI-written work as your own.

Why are non-users more worried about critical thinking?

RAND found 78% of students who avoid AI believe it harms thinking skills — possibly because they observe peers skipping the learning step entirely.

Which AI tools are safest for learning?

Tools designed for study — summaries, quizzes, flashcards, and guided explanation — tend to support retention better than open-ended chatbots that deliver finished answers.

Back to blog