The State of AI Fluency in Higher Education: 2026 Results

Summary

Higher education has stopped arguing about whether to use AI and started confronting how little anyone actually understands it.

That's the finding at the center of DataCamp's second annual AI Fluency Report, presented on a live webinar this week to walk through the 2026 results. The survey drew roughly 300 educators and students across 45 countries, mostly in higher education, and compared their answers against DataCamp's separate AI Literacy Report of business leaders. Adoption is close to universal: 71% of educators use AI daily and 93% weekly, while 90% of students use AI for coursework and 55% do so every day. ChatGPT, Google Gemini, and Claude lead the field, and Claude jumped from a marginal presence last year to third place this year. Usage has outpaced understanding, though. Eighty-nine percent of educators say AI fluency will matter to students' careers within five years, yet only 3% believe their own students are fluent now. Eighty-three percent of educators have already redesigned their assessments in response, favoring process-based work and supervised, in-person tasks over traditional no-AI assignments. Institutions have mostly sat this out: most have no formal AI governance policy, and 44% of students don't know if one exists. New this year, DataCamp compared its own platform data and found enterprise learners taking far more AI-related coursework than academic learners, a gap that widened rather than closed over the past year. Students say they'd join a formal AI fluency program if their school offered one.

Key Takeaways

• Educators and students have adopted AI at near-universal rates, but the report finds a wide gap between using AI and actually being fluent in it.
• Only 3% of educators believe their students are AI fluent today, even though 89% say fluency will be essential to their careers within five years.
• Claude went from barely registering in last year's survey to the third most-used AI tool among educators and students this year, behind ChatGPT and Google Gemini.
• Educators' top concern isn't cheating. It's that AI use is eroding students' critical thinking and foundational skills like programming and reading comprehension.
• Eighty-three percent of educators have redesigned their assessments for the AI era, almost entirely on their own initiative rather than under institutional direction.
• Students rate process-based assessments, like showing drafts and reasoning, as more valuable than traditional assignments that ban AI outright.
• Most institutions still have no formal AI governance policy, and 44% of students don't know whether their school has one at all.
• Only 2% of students report receiving any formal AI training from their institution.
• DataCamp's own platform data shows enterprise learners now put roughly 40% of their coursework toward AI fluency topics, compared with just 9% for academic learners, and that gap grew over the past year instead of closing.
• Eighty percent of students say they would join an institutional AI fluency program if one existed.

Deep Dives

The Gap Between Using AI and Understanding It

The adoption numbers in DataCamp's AI fluency report leave no room for debate: teachers and students in higher education have already gone all in on AI. Seventy-one percent of educators use it daily, 93% weekly. Ninety percent of students use it for coursework, and 55% do so every day. ChatGPT and Google Gemini remain the most common tools, but the real movement is Claude, which the report's presenter described as having gone from a name almost nobody mentioned last year to the third most-cited tool this year. "That's also where we have seen the market going as well," the presenter noted, pointing to the same pattern showing up in DataCamp's own AI curriculum search data.

What the numbers don't show is whether anyone is good at using these tools. That's the distinction the report draws between AI usage and AI fluency, which it defines precisely: "the capacity to collaborate productively and responsibly with AI systems through understanding their capabilities, limitations, and implications." Judged against that definition, the gap is stark. Eighty-nine percent of educators say AI fluency will be essential or very important to students' careers within five years. Only 3% think their students have it today.

That 86-point spread is the report's central tension: an entire generation of students has normalized AI as a daily tool, while the people teaching them see almost none of that translating into the judgment, skepticism, and technical understanding that would make it a genuine skill. The report treats fluency as a distinct, teachable competency, not a byproduct of exposure. Frequent use doesn't guarantee it, which is why the next question in the survey, and the next section of this report, was what's actually driving that gap.

Why Educators Are Rewriting Their Assessments

Ask educators what worries them about AI fluency and the answer isn't academic dishonesty, though that made the list. Their top concern is that students are becoming overreliant on AI and losing critical thinking skills in the process. Loss of foundational abilities, like programming and reading comprehension, ranked second, followed by an inability to evaluate AI output quality, then academic dishonesty, unequal access, and the risk of graduating without the AI skills employers actually want.

One anonymized quote from a social science faculty member, shared during the webinar, captured the concern in a single line: "inexperienced and novice learners lose most of the formative struggle of learning when they use AI tools." The presenter called it representative of how many educators think about the tradeoff.

Faced with that, 83% of educators have already redesigned their assessments for the AI era, and the presenter was blunt about where that initiative came from: "all of these assignments have been redesigned purely based on a professor's own initiative," not institutional guidance. When DataCamp asked students which redesigned formats they found most valuable, three answers stood out. Process-based work that shows drafts and reasoning ranked highest, followed by assessments that let students use AI openly and explain how they used it. In-person, supervised formats, like oral exams, came next. Traditional assignments that ban AI outright ranked lowest of all, a signal that students see them as disconnected from how they actually work now.

The pattern suggests educators have concluded that banning AI doesn't solve the fluency problem. Teaching students to show their reasoning, and to be transparent about where AI helped, does more to build the judgment the report says is missing.

Institutions Are Missing From Their Own AI Transition

If assessment redesign is happening from the ground up, institutional policy is nowhere to be found. When DataCamp asked professors and students how their institutions were responding to AI, the large majority reported no clear AI governance structure or policy of any kind. Forty-four percent of students said they didn't even know whether their institution had an AI policy, let alone what it said.

The presenter framed the disconnect directly: the changes happening on campus are "basically grassroots efforts between teachers and students," not decisions coming from institutional leadership. That absence extends to training as well as policy. Only 2% of students surveyed said they'd received any formal AI training from their school. Everything else, the report suggests, has come from students experimenting on their own or picking up habits informally, without structured guidance from the institutions responsible for their education.

That combination, no policy and almost no formal training, helps explain why fluency lags so far behind usage. Students are being asked to use judgment about tools that arrived faster than any institutional framework for teaching that judgment. Educators are filling the gap individually, redesigning assessments and setting informal expectations, but without institutional backing that guidance varies wildly from one classroom to the next. The report's authors treat this as the structural bottleneck behind the entire fluency gap: individual effort at the classroom level, absent any coordinated policy at the institutional level.

Classrooms Fall Behind the Workplace on AI Skills

New in this year's report, DataCamp compared its own platform data across two very different learner groups: enterprise employees taking DataCamp courses through their employer, and academic learners taking DataCamp courses through their school. The comparison quantifies a gap that educators had only been able to describe in general terms.

Of all the courses an enterprise learner takes on DataCamp, close to 40% are now related to AI fluency, up from around 16% just a year or two earlier. Academic learners, over the same period, put only about 9% of their coursework toward AI fluency topics, and that share didn't grow meaningfully between the 2025 and 2026 academic years. Enterprise learning nearly tripled its AI focus while academic learning stayed flat.

The presenter drew the implication out explicitly: "there's a huge gap between what the industry is learning, what employers are probably expecting of the graduating class, versus what the students are actually learning." It's one thing for educators to report concern about a fluency gap in surveys. It's another to see it show up in the actual courses students choose, or are assigned, compared with what employers are pushing their own workforces toward. The workforce isn't waiting for higher education to catch up, and the data suggests the distance between the two is growing rather than closing.

Closing the Gap: Upskilling, Policy, and a Pledge to Reach a Million Learners

Asked what would actually close the AI fluency gap, educators converged on one answer above the rest: comprehensive faculty upskilling, so teachers can set credible expectations for their students before asking students to meet them. Behind that, they pointed to updated AI-integrated curricula, AI fluency built into formal program learning outcomes, clearer institutional strategy and leadership, investment in tools and infrastructure, better policy and ethical coverage, and closer ties to industry.

The report ends on a more hopeful data point: students aren't the obstacle. Eighty percent said they would join an institutional AI fluency program if their school offered one. The presenter framed this as evidence against a common assumption, that students resist structure or oversight around AI use. "Students are willing to become AI fluent if the help was there," she said, arguing the barrier is institutional supply, not student demand.

DataCamp's own response is a pledge to upskill one million teachers and students worldwide during 2026, through DataCamp Classrooms, which is free to access at datacamp.com/universities and covers everything from basic data and AI literacy through business intelligence, data and machine learning engineering, and cloud development tools. DataCamp reached its first million Classroom users over six years, since 2019; this year's pledge is to do it again in twelve months. Certifications, including fundamentals like AI and data literacy, career tracks such as data analyst or AI engineer, and specialist credentials in data governance and the EU AI Act, are free for Classroom users.

For anyone trying to gauge their own AI fluency rather than wait for an institutional program, the presenter pointed to two options: Anthropic's AI Fluency Framework, the subject of the next webinar in DataCamp's series, and a free, roughly 15-minute AI fluency assessment on DataCamp that benchmarks a learner's skills against others who've taken it.