Results of the Fall 2025 U-M Undergraduate Student GenAI Sentiment Survey

More than 50% of students said they had moderate AI knowledge 

• 31% cited having a lot or a great deal of AI knowledge
• Only 15% of participants had little to no AI understanding
• More than 45% reported having a moderate amount of experience with AI
• 31% a lot or a great deal
• 23% of participants had little to no AI experience
   

U-M undergraduates indicated that they are navigating a patchwork of policies and approaches to AI in their courses, from not allowing AI at all or requiring explicit disclosure, to permission for select purposes, to no policy at all. 

• AI was not allowed for any purposes (61%)
• Any AI use for an assignment needed to be disclosed (61%)
• AI could be used for some processes but not whole assignments (47%)
• AI use was only permitted for some assignments (41%)
• AI was acceptable for tasks that were not turned in, but not for assignments (31%)
• There was no AI policy (25%)
• There was an AI policy, but it was not clear what the policy was (20%)
• AI use was permitted for all purposes (18%)
   

Among all surveyed students (n=1,611), those who reported using AI for writing, coding, or summarizing were asked additional questions to better understand their motivations and use patterns.

• Task Types and % of Respondents who use AI
  • Summarizing information (e.g., readings, notes, text, lectures, etc.) (68%)
  • Searching for information/research (68%)
  • Writing or editing text (63%)
  • Solving quantitative or mathematical problems (60%)
  • Planning for or outlining work tasks (44%)
  • Personal or work life management (e.g., scheduling, routine tasks) (41%)
  • Coding and programming assistance (33%)
  • Generating creative content (e.g., images, video, audio) (30%)
  • Something else (13%)

AI in Writing

• 799 students (63%) use AI as part of their writing and editing process. These undergraduate students report using AI for a variety of tasks, including:
  • Editing at the End
    • 80% of students use AI to proofread and copy edit their work at the end of the writing process.
  • Brainstorming
    • 77% of students use AI to generate ideas and brainstorm.
  • Turning Ideas into an Outline
    • 66% of students use AI to transform their ideas into an organized outline or structure.
  • Producing Text
    • 28% of students use AI to generate a first draft and 14% have AI produce an entire text.

AI a Study Tool for Summarizing Information

• Among the 871 students (68%) who reported using AI to summarize information were asked how they apply it across different academic tasks such as:
  • Study Methods and Comprehension
    • Most students (71%) use AI to summarize assigned readings and study materials, highlighting how common it is used as a study tool.
  • Support for Prep and Research
    • Almost half of those using AI to summarize use these tools to condense research papers or lecture notes (40-45%).
  • Less focus on Writing Revisions
    • Only 37% use AI to summarize their own writing, indicating it’s applied more for comprehension than personal editing.

AI for coding

• 426 students (33%) who reported using AI for coding and programming assistance were asked how they apply it across different academic tasks.
• Providing Feedback and Editing
  • The majority of students use AI (65%) to diagnose bugs in their code and brainstorm (46%)
    alternative code methods
• Drafting and Completing Code
  • 37% of these students utilize AI to produce their first coding drafts, and only 29% use AI to produce full code outputs.

Key Motivations Behind AI Use

• AI Saves Students Time
  • Over two-thirds of students use AI in writing, summarization, and coding because it saves time. Specifically, 66% of those who use AI for writing, 77% for summarization, and 66% for coding identify time efficiency as a primary motivation.
• Provides Feedback and Support
  • Receiving feedback was viewed as a key benefit among 47% of the students who use AI to code, 30% who use AI for summarization, and 72% who use AI for writing.
• Completing an Entire Task
  • In contrast, a smaller share of students use AI to fully generate outputs. 28% use AI to produce a first written draft, and 14% to create entire texts.
  • In coding, 37% use AI to generate initial drafts, while 29% rely on it to complete full code outputs.
     

Overestimates of Peer Use and Misuse of AI

One of the most noteworthy findings about student perception is that they believe their own AI use and likelihood of policy violation is far lower than that of their peers.

• 20% of students say they always or usually use AI, while 74% believe their peers do use AI that frequently. 
• 78% of students report always or usually adhering to AI policies, and only 6% of participants rarely or never follow AI policies. In contrast, 80% of respondents perceive their peers as never, rarely, or sometimes following AI policies. Only 2% of students believe other students always follow AI policies.

Although it is conceivable these results reflect a social desirability response (individuals underreport practices seen as negative), it is equally possible that the phenomenon is the same that leads students to believe their peers drink more alcohol than they do (e.g., see this article by Perkins, Haines, & Rice, 2005).

Perceived Sufficiency of Support for GenAI

There is also a gap between student perception of the amount of support they believe U-M offers to students around AI vs what they would prefer:

• 70% of students believe that Michigan is doing little to nothing at all to prepare students to use AI effectively
• 85% believe the University should provide moderate to extensive preparation.