AI Emergency at College: How 90% of Students Use ChatGPT Daily

Nearly nine in ten university students worldwide now use artificial intelligence tools for their studies, with ChatGPT remaining the dominant platform, according to the Digital Education Council Global AI Student Survey 2024, whose updated 2026 dataset was released on 2 June. The study, which polled more than 3,800 students across 16 countries, found that 24% of respondents use generative AI on a daily basis and a further 54% use it at least weekly. Faculty bodies in the United States, the United Kingdom and across the European Higher Education Area are racing to adapt assessment models that were designed for a pre-ChatGPT classroom.

The pace of adoption has outstripped institutional response. Universities that issued cautious «wait-and-see» memos in 2023 now confront cohorts in which AI literacy is the norm rather than the exception. The result is a regulatory vacuum in which professors improvise policy module by module, students self-report rampant uncertainty about what is permitted, and the integrity of degrees granted between 2024 and 2026 has become a live debate in academic senates from Stanford to the Sorbonne.

The college ChatGPT use rates that triggered the alarm

Multiple peer-reviewed and institutional studies converge on the same conclusion: ChatGPT use is no longer a fringe behaviour but the modal student workflow. The Higher Education Policy Institute’s 2025 survey of 1,041 UK undergraduates found that 92% had used generative AI for academic tasks, with 88% specifically citing assessments — a near-vertical climb from 53% the year before.

The composition of that usage matters. Students no longer reach for AI only to polish prose. The HEPI data shows 18% of respondents admit to including AI-generated text directly in submitted work, while 58% use the tools to explain concepts and 53% to summarise reading lists. In the United States, an EDUCAUSE 2025 student survey reported that 59% of undergraduates use AI weekly for coursework, with the steepest growth among first-year students who arrived on campus already fluent in prompt design from secondary school.

Adoption is uneven by discipline. STEM students lean on AI for code generation and problem decomposition. Humanities students use it predominantly for outlining, translation and editorial polishing. Business and law schools, where written argumentation drives grading, are now the loudest constituency pushing for oral and in-person assessment formats.

Why ChatGPT use among university students is reshaping assessment

The challenge for universities is no longer whether to acknowledge AI but how to grade in its presence. According to the Russell Group’s joint statement on generative AI, updated in February 2026, all 24 member universities have replaced blanket bans with discipline-specific policies that distinguish «AI-assisted» from «AI-generated» work, but enforcement mechanisms remain inconsistent across departments.

Detection technology has so far failed to close the gap. OpenAI itself discontinued its AI Text Classifier in 2023 citing low accuracy. Turnitin’s AI writing indicator, deployed across thousands of institutions, has been the subject of bias complaints from non-native English speakers, whose simpler syntactic patterns produce false positives. A 2024 study published in International Journal for Educational Integrity found Turnitin’s tool flagged human-written essays by ESL students as AI-generated at roughly twice the rate of native-speaker submissions.

The pedagogical response has split into two camps. The first treats AI as a calculator-equivalent tool whose use must be disclosed and which complements rather than replaces learning outcomes. The second, championed by a vocal minority of faculty, calls for a wholesale return to in-person, handwritten, time-limited assessment — what one Oxford tutor described to Times Higher Education as «the only invigilation model that survives contact with ChatGPT».

«The horse has bolted. Pretending that we can detect our way back to pre-2023 essay grading is a fantasy. Our obligation is to redesign assessment so that AI use makes the answer better, not so that AI use makes the answer possible.»

What students actually do with ChatGPT

Student behaviour is more nuanced than headline figures suggest. The Digital Education Council survey identifies four dominant use cases — explanation, summarisation, brainstorming and editing — each used by more than 40% of respondents. Pure substitution, where AI writes the submitted text, is reported by a smaller but meaningful 18% to 22% of students depending on the study.

The table below compares the most cited use cases across three recent international surveys.

Students also report a productivity gap they cannot ignore. In the DEC dataset, 80% said AI tools have improved the quality of their work, and 73% said they would feel disadvantaged not using them while peers do. That collective-action dynamic — the so-called «AI prisoner’s dilemma» of higher education — explains why university ChatGPT use rates rose by more than 30 percentage points in twelve months in some jurisdictions.

Public policy is starting to catch up. The European Commission’s Digital Education Action Plan, updated in early 2026, now references generative AI explicitly. National-level guidance is also accelerating across OECD countries; the United States’ Department of Education AI Policy Brief and the United Kingdom’s Office for Students framework both arrived in 2025. Readers tracking that policy layer may find StudyVerso’s earlier explainer on the government AI push in schools a useful primer on how regulators are framing classroom AI for the K-12 pipeline that feeds these universities.

The literacy gap behind the headline numbers

While usage is near-universal, formal training is not. The Digital Education Council reports that 58% of students believe they lack sufficient AI knowledge and skills, and only 36% have received structured AI literacy instruction from their institution. The mismatch produces graduates who use AI confidently but cannot always evaluate when its output is wrong.

Faculty surveys show a parallel gap on the other side of the lectern. A 2025 Tyton Partners study of 1,800 US faculty found that only 22% had received formal training on AI in teaching, even though 60% said they had changed at least one course design element in response to it. The result is a teaching workforce learning the tools at the same time as the students it is supposed to be assessing.

Hallucination remains the most cited risk. ChatGPT, Claude and Gemini all continue to fabricate citations and misattribute quotations, and the volume of retracted student work containing invented sources is rising. The University of California academic integrity offices reported a doubling of cases between 2023 and 2025, with fabricated bibliography the single largest sub-category.

Implications for students, faculty and the degree itself

The stakes extend beyond grade integrity. Employers in consulting, law and finance have begun asking whether 2024–2026 graduates can write a structured argument without AI scaffolding, a concern reflected in McKinsey’s 2025 Future of Work report, which highlighted «unmediated writing» as a screening criterion at several Fortune 500 firms.

For students, the practical implication is that disclosure and method literacy now matter as much as the output itself. Institutions that have adopted «AI use statements» — a short paragraph in which students declare which tools were used and how — report higher trust and fewer formal misconduct cases. The University of Sydney’s two-lane assessment model, splitting work into AI-permitted and AI-secure tasks, has been cited as a template by at least nine other universities in the past year.

For faculty, the shift means that the marginal value of grading lies less in catching cheats and more in designing tasks where AI use is visible, productive and defensible. EdTech startups including Modo Cheto, Quizlet and Khan Academy’s Khanmigo are pitching themselves to universities as classroom-grade AI tutors with audit trails, although independent evaluation of their pedagogical claims remains thin. Readers comparing the broader category may also want to revisit the analysis of how public bodies are framing AI tools for younger learners, where the same vendors are competing for procurement.

For the degree, the questions are existential. If 86% of students use AI and only 36% have been trained in it, the implicit certification carried by a diploma — that the holder can independently produce work of a given standard — comes under pressure. The next academic year, with the first full graduating cohort whose entire undergraduate experience coincided with ChatGPT availability, will be the first real-world test of whether universities have adapted in time.

The open question, as the 2026–2027 academic year approaches, is not whether students will continue to use ChatGPT daily — they will — but whether universities will treat that fact as a crisis to police or as a pedagogical reality to design around. The answer will define what a degree means for the next decade.