If your graduate hiring framework is built around the assumption that incoming cohorts have used generative AI selectively or occasionally during their studies, the 2025 Chegg data shows the assumption needs replacement. 80% of university students across 15 surveyed countries say they have used generative AI to support their learning. The 2023 figure was 40%. The doubling happened in 24 months. 56% of users input a question at least once a day. The cohort entering the workforce in 2026 is now AI-native by default, not by exception.

University student GenAI use by country, 2023 vs 2025:

The country-level data shows the universality. Indonesia leads at 95%. Malaysia and Saudi Arabia at 89% and 87%. India at 83%. Across the 15 surveyed countries, 14 are above 67%. Only the United States, the United Kingdom, and Turkey are at the 67% level, meaning the floor of the distribution is at two-thirds of students using AI for schoolwork. The country variation is meaningful (28 percentage points between top and bottom) but the cluster is tight: everywhere is above two-thirds.

The growth pattern is also worth noting. The countries with the largest 2023→2025 increases are South Africa (+61pp), Saudi Arabia (+50pp), South Korea (+48pp), the US (+47pp), and the UK (+48pp). The countries that started lower have caught up rapidly. The distribution is converging on a common high-use baseline rather than splitting into adopters and non-adopters.

University student GenAI uses for schoolwork, 2025:

The use-case data is the other side of the structural picture. The top university student GenAI use is "understanding a concept or subject" at 56%. Then "researching for assignments and projects" at 52%. "Generating initial ideas/first drafts" at 46%. "Writing/editing assignments and essays" at 41%. The pattern is the use of GenAI for cognitive scaffolding (explanation, ideation, drafting) not for substitution of cognitive work.

This shifts the conversation about graduate AI use. The framing that has dominated K-12 and early university discourse, "students will use AI to cheat on assignments," is not the dominant use pattern in the data. The dominant use is conceptual support during learning. The implication: the cohort entering the workforce has spent multiple years using AI as a learning aid, not as a shortcut around learning. They have built habits of using AI in a particular way during their formative academic years.

Three structural implications follow for graduate hiring frameworks.

The first: the assumption that "incoming graduates can use AI" is now structurally correct rather than aspirational. Every recent graduate cohort from 2026 forward will have at least two years of consistent GenAI use as part of their academic workflow. The capability question is not "can they use AI tools" but "how have they been using AI tools, and does that match how the organisation wants them to use AI tools." Hiring assessment frameworks designed pre-2024 that test for "AI literacy" as a discriminator may produce no discrimination: every candidate now meets the baseline.

The second: the use-pattern differentiation has become the relevant discriminator. The structural data shows the majority of students use AI for conceptual support; a meaningful minority use it for direct task substitution. A subset use it for advanced applications (writing code, building agents, interfacing with technical systems). The differentiator at hiring is not "does this candidate use AI" but "what is the depth and shape of this candidate's AI use." Hiring frameworks that probe specific use cases, such as "show me how you've used AI in a research project," will surface the distribution that the binary "do you use AI" question now collapses.

The third: the organisational AI use pattern needs to match (or rationally diverge from) the graduate cohort's use pattern. If the graduate's academic AI workflow is conceptual scaffolding and the organisation wants AI use in workflow automation, there is a transition cost: the graduate is bringing a use pattern that doesn't match the role. This isn't a problem in itself, but it is a planning variable. The 24-36-month transition window for new graduates includes adapting their AI use patterns from academic to organisational contexts. Organisations that recognise and design for this transition will integrate graduates more smoothly than organisations that assume their academic AI use directly transfers to organisational AI use.

What does the country variation mean for global hiring? The 2025 numbers suggest that incoming graduates from Indonesia, Malaysia, Saudi Arabia, India, and other high-use countries are statistically more likely to have extensive GenAI use experience than incoming graduates from the US, UK, or Turkey. The difference is 25-30 percentage points at the country level. For organisations hiring global talent pools, the country-level AI use baseline is now a meaningful (if imperfect) signal in candidate screening.

The data observation: the academic AI-use baseline globally is 80% and rising. The 2025 number doubled in 24 months from the 2023 baseline. Whether the curve plateaus in 2026-2027 or continues upward toward the high-90s is open. What is settled is that the cohort entering the workforce from 2026 onward has spent the majority of their academic life with consistent GenAI use as a learning aid, and graduate hiring frameworks calibrated against a pre-2024 baseline are reading a stale signal.

For workforce strategy teams sizing AI capability against graduate cohort assumptions, the planning anchor needs to shift. The graduate cohort is no longer the marginal AI adopter inside the organisation. They are the most experienced AI users for the longest period. The organisational AI strategy needs to absorb that asymmetry. Strategy that assumes the organisation will teach graduates AI tools has the polarity reversed; in many domains the graduates will be teaching the organisation use patterns the organisation has not yet developed. Plans that recognise the polarity will adapt their onboarding and capability frameworks accordingly. Plans that don't will continue to operate against the wrong direction of knowledge transfer.

Sources