92% of Students Use AI to Study: What It Means for Computing Applicants

Yes, the number is real. According to the Higher Education Policy Institute's 2025 Student Generative AI Survey, 92% of UK undergraduates now use AI in some form to study, up from 66% just a year earlier. HEPI's follow-up 2026 survey puts the figure even higher, at 95%. For most students that is simply how coursework gets done now. For anyone applying to a computing degree specifically, it changes something bigger: the subject itself is the one producing the tools driving this shift, which means the skills, admissions expectations, and career value of a computing degree are all being renegotiated at the same time.

This matters more for computing applicants than for almost any other subject choice. A history student using AI to summarise reading lists is adapting to a general-purpose tool. A prospective computer scientist is choosing to study the exact discipline that builds those tools, at the exact moment employers, universities, and exam boards are still working out what a computing education should look like when a lot of routine code can already write itself. This guide walks through the data, what it means for admissions and assessment, and how Academic Training Hub (ATH) structures its NCC Education computing pathway around that reality rather than pretending it isn't happening.

How Many Students Actually Use AI to Study? The 2024-2026 Numbers

Student AI use has gone from a minority habit to a near-universal one in under three years. HEPI's annual survey, run with Kortext and the polling firm Savanta, is the most consistently cited UK data source for this trend, and its year-on-year jump is the sharpest recorded shift in recent higher education research.

The 2026 report, based on 1,054 full-time UK undergraduates surveyed by Savanta in December 2025, describes AI use as having become "near universal." Its own framing is blunt: the meaningful question is no longer whether students use AI, but how well they use it. That reframing is exactly why a page titled "is a computing degree still worth it" needs a different answer today than it would have needed three years ago.

Read those five rows together and a pattern appears that a single statistic hides: students lean on AI heavily for process (research, structuring, checking) but the share who paste AI output directly into graded work, while still small, is growing every year. That is precisely the behaviour universities are now designing assessment to catch and discourage.

Why This Shift Hits Computing Applicants Differently

Computing is the one subject where the tool causing this disruption and the subject matter being taught are the same thing. A computing student is not just using AI to study, they are studying the field that builds the AI they are using, which is why the numbers above land differently for this audience than for any other applicant group.

Three consequences follow directly from that gap, and each one affects how a prospective computing student should evaluate a course:

• Fundamentals matter more, not less. If AI-suggested code cannot always be trusted, the person reviewing it needs to understand data structures, algorithms, and system design well enough to spot when it is wrong.
• Assessment is shifting toward process, not just output. Courses are increasingly testing how a student reasons through a problem, not only whether the final code compiles, because a working answer no longer proves the student understood it.
• AI literacy is becoming a taught skill, not an assumed one. Knowing how to prompt, verify, and correct an AI tool is now treated as a professional competency in its own right, the same way debugging or version control is.

Fewer Students Are Choosing Computer Science, More Are Choosing AI Degrees

UK computing degree acceptances fell in the 2025 admissions cycle, but not evenly, and the pattern tells a clearer story than the headline number does. Analysis by BCS, The Chartered Institute for IT, of UCAS's 2025 end-of-cycle data found that 31,670 students of all ages were accepted onto UK computing degrees, a 3% drop, though still the second-highest total on record. For 18-year-old applicants specifically, acceptances fell 9% to 16,865, the first decline in seven years and a sharper drop than the wider undergraduate market, which grew 3.5% over the same period.

There is a genuine silver lining in the same dataset. The gender gap in UK computing degrees narrowed to its smallest ratio on record, 4 to 1 male to female, down from 5.5 to 1 in 2019/20, and half of all computing students in 2025 came from disadvantaged backgrounds, compared with 41% across all university subjects. A subject going through this much change is also, by the same data, becoming more accessible to more people, not less.

Is a Computing Degree Still Worth It If AI Can Write Code?

Yes, but the value has moved, not disappeared. A computing degree is not primarily proof that someone can produce working code, since AI tools can already do a meaningful share of that on their own. Its real value now is proof that someone can specify a problem correctly, evaluate whether AI-generated code is actually correct and secure, and design a system AI could not have assembled unsupervised.

Put plainly, a computing degree earned in 2026 is being marked against a different rubric than one earned in 2019. The following comparison reflects that shift as it is actually playing out in curricula and hiring, not a hypothetical future version of it.

How Admissions and Assessment Are Changing Because of AI

Universities have not waited for a settled policy consensus before acting. HEPI's 2026 data shows that nearly two-thirds of students report their assessment has changed noticeably because of AI, and the same report tracks a steady rise in students submitting AI-generated text directly, from 3% in 2024 to 12% in 2026. Institutions are responding with more oral defences of written work, in-person practical exams, and coursework that asks students to document their reasoning process rather than only their final answer.

The applicant-side picture is shifting too. Research summarised by Wonkhe, drawing on UCAS applicant data, found that 48% of prospective students used AI tools to research their options, with 61% comparing universities and 59% exploring subject choices through AI. But trust drops sharply once decisions get personal: only 13% said they trusted AI more than a teacher, and 73% had already encountered inaccurate information from an AI tool during their research. Applicants are using AI to widen the funnel of options, then relying on people, not algorithms, to narrow it down.

"The question is no longer whether students use AI, but how well they use it." HEPI, Student Generative AI Survey 2026

What Employers Actually Want From Computing Graduates Now

Employers want graduates who make AI tools more useful, not graduates who compete with them at the same narrow task. That is the consistent thread across the WEF's skills data and the BCS enrolment figures cited earlier: growth is concentrated in AI-adjacent specialisms, and analytical thinking outranks every specific technical skill in employer priority. In practice, that means a computing graduate stands out for:

• System-level design, because AI tools remain far more reliable at generating a function than architecting an entire application.
• Verification and testing discipline, since someone has to catch the errors AI-generated code still produces.
• Security and ethical judgement, an area BCS and WEF both flag as growing faster than general software roles.
• Clear technical communication, needed to specify a problem precisely enough for an AI tool, a teammate, or a client to act on it correctly.
• Independent project delivery, the exact skill a final-year capstone project, like the one built into ATH's Level 6 BSc (Hons) Computing top-up, is designed to prove.

How to Use AI Without Sabotaging Your Own Learning

The safest rule for a computing student is simple: use AI to check your understanding, never to replace building it. HEPI's own data shows the students most at risk are the 12% pasting AI output straight into graded work, since that habit produces a degree certificate without the underlying skill an employer is actually paying for. A short, practical checklist keeps AI useful without hollowing out the learning it is meant to support:

• Write your own first attempt before asking AI for help, so you know where your understanding actually breaks down.
• Ask AI to explain its reasoning, not just produce an answer, and check that reasoning against what you have been taught.
• Never submit AI-generated text or code without rewriting it in your own words or your own logic first.
• Keep a note of which tools you used and how, since more UK universities now require this disclosure as standard practice.
• Use AI to generate practice questions or edge cases for your own code, not to generate the code itself.
• Treat every AI answer as a first draft from a fast but unreliable collaborator, not as a finished, trustworthy result.

Choosing the Right Computing Pathway in an AI-Driven Job Market

The staged NCC Education route lets a student build AI literacy gradually, alongside the foundational skills that make it possible to use AI critically rather than blindly. Each stage adds a layer that matters more, not less, now that AI tools are part of everyday practice.

Anyone weighing this decision in more general terms, including whether a Level 6 qualification carries the same weight as a full honours degree, can find that comparison worked through in detail in ATH's guide to whether a Level 6 qualification is equivalent to a degree.

Common Mistakes Applicants Make When Judging AI's Impact on Computing Study

Most of the confusion around this topic comes from a handful of recurring assumptions, each of which the data above directly contradicts.

• Assuming AI makes a computing degree pointless. Enrolment data shows the opposite: applicants are moving toward AI-integrated computing degrees, not away from computing altogether.
• Assuming high AI usage means high trust. Developers use AI tools constantly while distrusting their accuracy more than they trust it, which is exactly why human verification skills remain the priority.
• Treating AI-detection as the whole academic integrity story. Universities are moving toward process-based assessment, oral defences, and staged submissions, not just detection software, so understanding your own work matters more than avoiding a checker.
• Ignoring how much this varies by computing specialism. Traditional Computer Science enrolments fell 10% in 2025 while dedicated AI degrees rose 42%, so the safest choice is rarely the most generic-sounding title.
• Believing a foundation-to-degree pathway is a shortcut around rigour. A staged route through NCC Education Levels 4, 5 and 6 covers the same 360 credits and the same honours-level final project as three years studied entirely in the UK.

The honest answer to "is a computing degree still worth it" is that the degree itself has not lost value, but the version of it worth having has changed. A course that still measures success purely by whether the code compiles is teaching to a world that is already gone. A course built around judgement, system design, and the discipline to check an AI tool's work, the way ATH's NCC Education pathway is structured, is teaching to the one employers are actually hiring into.

Frequently Asked Questions