The AI Challenge: How college faculty assess the present and future of higher education in the age of AI

Synthesis: A national survey of 1,057 US college and university faculty, conducted by the American Association of Colleges and Universities with Elon University's Imagining the Digital Future Center between 29 October and 26 November 2025, records what instructors think generative AI is doing to their teaching and their students. The headline finding is not resistance to the technology but a split verdict on it: 95% expect generative AI to increase students' over-reliance on AI tools, 94% expect it to increase academic integrity concerns, 90% expect it to diminish critical thinking, and 83% expect shorter attention spans, while 61% still expect it to improve and customise learning. Almost three-quarters (73%) say they have personally handled academic integrity cases involving generative AI. Faculty have written their own rules far faster than their institutions have — 87% set their own assignment-level policies, against 35% who say their department has written guidelines and 48% who say their institution has. The report is a non-scientific sample that its authors explicitly say is not generalisable, so it is best read as the sector's expressed concerns rather than as measured effects.

Key Findings

1. Cheating is seen as up, and nearly everyone has handled a case. 78% said cheating on their campus has increased since generative AI tools became widely available, including 57% who said it increased a lot, while 18% did not know. 33% reported having a lot of academic integrity cases in their own courses and a further 40% at least a few; 73% said they had personally dealt with integrity issues involving students' generative AI use.
2. Institutions are judged unprepared. 59% said their schools are not very or not at all prepared to use generative AI effectively for preparing students for the future, 68% said their schools have not prepared faculty to use it for teaching and mentoring, and a similar share for scholarship, with 57% and 55% reporting the same for non-faculty and for staff in institutional operations such as recruitment, student life, athletics, fundraising and alumni relations.
3. Graduates are judged unready. Asked about the previous spring's graduates, 63% said they were not very or not at all prepared to use generative AI in the world of work, 62% said they were unprepared in their overall understanding and use of the tools, and 71% said they were unprepared in their understanding of the ethical issues those systems raise.
4. A substantial minority opts out. 26% of respondents said they do not use generative AI tools at all, and a third said they choose not to use them for teaching. Non-use is unevenly distributed: 40% of arts and humanities faculty and 28% of social scientists reported not using the tools. Faculty see each other as the obstacle — 82% called faculty resistance a challenge to departmental adoption and 83% cited unfamiliarity.
5. There is little consensus on what counts as cheating. 52% said it is cheating for a student to follow a detailed AI-generated outline when writing a paper, leaving 47% who called it legitimate or were unsure. 45% considered it legitimate for a student to write a paper and then make the edits a system recommends, while 55% called that illegitimate or were unsure. On faculty's own use, respondents were split on using generative AI for a first-draft syllabus, for slide decks, and for replies to student emails, but clearer that using it to grade essays or to write portions of a submitted journal article is not legitimate.
6. AI literacy is valued by half and taught by more. 49% said it is extremely or very important for students to develop AI literacy skills before graduating, against 13% who called those skills irrelevant and 11% who called them only slightly important. 69% said they have addressed AI literacy in their instruction, and large majorities said it is necessary to teach the associated issues — bias, hallucination, misinformation and deepfakes, Privacy, cybersecurity and environmental cost.
7. Rules are individual, AI Governance is thin. 87% have created their own policies for students on how they may and may not use generative AI, but only 35% said their department has written guidelines and 48% said their institution has. Structurally, 55% reported a task force or oversight group, 37% new AI-focused classes, 17% an AI major or minor, 16% new academic leadership offices, and 13% adoption of AI literacy as a general education outcome.

How the survey was conducted

An invitation was sent on 29 October 2025 to faculty known to AAC&U and Elon University, and the survey closed on 26 November 2025; 1,057 higher-education teachers responded to at least part of it. The authors describe the sample as diverse in academic discipline, size of undergraduate population and faculty status, and state plainly that it is a non-scientific sample whose results are not generalisable. Open-ended responses supply the quotations printed throughout the report. This is the second survey in the series, following an earlier canvassing of higher education leaders, which makes the comparison one of role rather than of time.

What faculty expect the technology to do

The report's most useful contribution is the balance sheet faculty constructed for the future, because it separates hope from fear within the same respondents rather than across camps.

• Hopeful: 61% said generative AI will improve and customise learning, against 32% who expect little or no impact. 41% expect improved research skills and 40% expect students to write more clearly and persuasively, though 53% and 58% respectively expect little or no impact on those dimensions.
• Fearful: 95% expect increased over-reliance (with 75% saying a lot), 94% increased academic integrity concerns (76% a lot), 90% diminished critical thinking (66% a lot), 83% shorter attention spans (62% a lot) and 81% wider digital inequities (58% a lot).
• Creativity is the negative expectation: 70% expect generative AI to affect students' creativity not much or not at all, while only 27% expect an increase.
• Careers: 49% said increased use of the tools in their field over five years will have a more negative than positive effect on students' careers, with 20% expecting the opposite, 20% expecting both equally and 11% unsure; 47% feared the long-term employment impact in their disciplines would be very or somewhat negative against 25% optimistic.
• The profession: 39% believe generative AI tools will diminish the role of faculty, and 86% think the impact of the tools on those who teach will be significant and transformative or at least noticeable, with only 4% expecting it not to amount to much.

Why this belongs next to the knowledge base's other evidence

The survey measures perception at scale, where most of the knowledge base measures behaviour or outcomes, and the two readings diverge in an instructive way. Faculty report that cheating has risen and expect critical thinking to fall, but the causal evidence on those claims is neither uniform nor as stark: the guardrail field trial showed assisted practice gains of 48% alongside an unassisted exam penalty of 17% that a guardrailed tutor removed entirely, and the performance-versus-learning distinction is what makes an instructor's impression of decline hard to verify from the artefact alone. The 95% expectation of over-reliance is a forecast about a mechanism the knowledge base describes directly as cognitive offloading and metacognitive disengagement, and the 26% who abstain are the population behind the faculty AI competency gap that institutional support programmes exist to close.

Two governance facts are the most actionable in the report. The first is the 87% to 48% gap between individual policies and institutional ones, which is the mirror image of what policy studies find at the institutional level and explains why students meet inconsistent rules in the same institution. The second is the thinness of the structural response — a task force in 55% of cases, but an AI literacy general education outcome in only 13% — which is the environment in which educational policy on AI has to be implemented, and the reason the knowledge base treats the support ecosystem, not the policy document, as the deliverable.

Limitations

The sample is self-selected from faculty known to AAC&U, so it over-represents engaged and possibly more AI-aware instructors, and the authors state that results are not generalisable. All measures are self-reported attitudes and estimates, including the cheating trend and the case-handling figures, which cannot be validated against institutional records. Several items were answered by people who had already said they do not use the tools, and the report marks those segments rather than excluding them. The forecast items are predictions about the next five years and therefore untestable at the time of publication.

Connected Concepts

• Academic Integrity — the concern that dominates the survey's negative expectations
• AI Literacy — valued by half the respondents, taught by more
• Critical Thinking — the outcome faculty most expect to decline
• Cognitive Offloading — the mechanism behind the over-reliance expectation
• Metacognition — the self-regulation the forecasts assume will weaken
• Student-AI Interaction — how students use the tools faculty are judging
• AI Detection — the response implied by integrity concern and questioned elsewhere
• Assessment Validity — what an assignment still measures once AI can complete it
• Teacher AI Competency — the capability gap the preparedness items describe
• Educational AI Policy — where the 87%/48% policy gap sits
• AI Governance — task forces, leadership offices and general education outcomes
• Change Management — adoption resisted by colleagues, not policy
• Trust — faculty scepticism about tools and about institutional readiness
• Equity — the digital-inequity expectation
• AI Misuse and Learning Harm — the harm the forecasts name
• Student Experience — readiness, wellbeing and career prospects as faculty see them

Connected Articles

• From fear to innovation: A case study of transformative faculty development for ethical AI integration in higher education — What a structured faculty development programme did to exactly these anxieties
• Governing generative AI in higher education: Emerging policy approaches and support ecosystems at innovative U.S. universities — Institutional policy and support ecosystems at innovative universities
• Governing academic integrity: Ensuring the authenticity of higher thinking in the era of generative artificial intelligence — An indicator framework for governors facing the integrity question
• The AI guilt complex: Moral emotions and ethical dilemmas in academic technology adoption — Moral emotions among academics adopting AI, including guilt about their own use
• Tools facilitate cheating, or partner supports learning? GenAI and academic integrity issues from students' perspectives — Students' own reading of the partner-versus-cheating question
• The Integrity of Psychology Assessments in the AI Age: A Critical Examination — What happens to a whole programme's assessment when the pass boundary is the operative variable
• Generative AI without guardrails can harm learning: Evidence from high school mathematics — The causal evidence behind the over-reliance concern
• Distinguishing performance gains from learning when using generative AI — Why an instructor's impression of decline is hard to verify

Citation

Watson, C. E., & Rainie, L. (2026). The AI challenge: How college faculty assess the present and future of higher education in the age of AI. American Association of Colleges and Universities and Elon University Imagining the Digital Future Center.