Generative artificial intelligence is forcing universities to reconsider what a degree should certify just as graduates confront a more difficult transition into work. The central question is no longer simply whether students have used AI, but whether they can judge its output, explain their decisions and produce work that demonstrates independent understanding.
The pressure is coming from both classrooms and employers. A 2025 report cited by CNBC found that 51% of employers considered the job market for that year’s college seniors poor or fair, the highest proportion since 2020–21. The same report said entry-level job postings in the United States had fallen 35% since January 2023, with artificial intelligence identified as one factor in the decline. CNBC reported.
Other findings point to a widening gap between academic credentials and early-career opportunity. Research from education company Cengage Group, reported by The Edu Ledger, found that 30% of 2025 graduates had secured full-time employment in their field, down from 41% for the previous class. The report also said that 33% of graduates remained unemployed and actively seeking work, while 26% had taken jobs unrelated to their studies. The Edu Ledger reported.
From bans to assessment redesign
Universities have responded unevenly. Some have tightened rules around unauthorised assistance, while others are redesigning assessment so that AI use becomes visible and discussable. The shift reflects a practical problem: detection systems can produce false positives, while take-home essays may reveal less about a student’s own reasoning when text-generation tools are readily available.
A UNESCO survey published in September found that nearly two-thirds of higher-education institutions connected to its Chairs or UNITWIN networks either had guidance on AI or were developing it. Yet only 19% reported having a formal policy; 42% said their frameworks were still under development. Institutions described measures including student guidance, mandatory AI literacy and redesigned assessment. UNESCO published the survey findings.
The University of Leicester offers one example of the emerging model. Its policy says that, from the 2025–26 academic year, every undergraduate programme should include at least one assessment requiring students to use generative AI, either by producing material with it or critiquing AI-generated content. The policy also says marking should not normally be undertaken by AI. The university’s policy states.
That approach has advantages and risks. Requiring students to interrogate AI can build skills relevant to workplaces where drafting, coding, research and analysis are increasingly augmented by software. But it also assumes equal access to effective tools and reliable internet connections. Students may face different paid-tool limits, privacy terms and levels of prior exposure, potentially turning assessment into a test of resources as much as learning.
What employers may value next
The graduate labour market is not being transformed by AI alone. Employers cited economic conditions and a tight labour market alongside adoption of the technology, according to the Cengage findings. Degree requirements may also be growing rather than disappearing: 71% of employers in that survey required a two- or four-year degree for entry-level roles, up from 55% in 2024. The Edu Ledger reported.
For students, that creates a paradox. A degree remains an entry ticket for many jobs, but the qualification alone may be less persuasive than evidence of applied ability. The same survey found that graduates ranked personal referrals, internships and prior work experience as more decisive in securing employment than the degree itself. That raises concerns about unequal access: unpaid placements, informal networks and professional mentoring can advantage students with greater financial and social support.
UNESCO’s guidance frames the alternative as human-centred AI education. Its student competency framework emphasises agency, ethics, technical understanding and the ability to assess AI systems rather than simply operate them. UNESCO described the framework.
The future of the university degree will therefore depend partly on whether institutions can connect academic learning with credible demonstrations of judgment. Students will still need disciplinary knowledge, but employers are likely to scrutinise how they solve unfamiliar problems, verify information, communicate clearly and use automation without surrendering responsibility.
That is a demanding brief for universities already managing stretched staff and competing expectations. A fair transition will require transparent rules, trained teachers, accessible tools and more paid routes into professional experience. Without those safeguards, AI could deepen existing inequalities. Used deliberately, however, it could push assessment closer to the capabilities graduates need when the first rung of the career ladder is becoming harder to reach.
Sources: CNBC; The Edu Ledger; UNESCO; University of Leicester.