Now that's what I call The Secret Life of Students 2026

Every year at The Secret Life of Students we seek to uncover a hidden truth about the student experience.

In the past we’ve looked at belonging, student health and students getting better. This year we tackled one of the sector’s biggest disruptors – that isn’t financial sustainability – artificial intelligence. But when we logged onto our weekly planning meetings it quickly became apparent we were not just talking about AI, digital tools or students “cheating,” we were actually talking about students, their learning and the role of assessment.

For all the challenges that AI brings, it does force us to get (un)comfortable with the first principles of learning and reassess what learning is looking like in 2026 and perhaps what it should look like.

Across the day we were envisioning a student experience that is more human, not less. We were thinking about how students are working the system today, what employers value in graduates, the intersection of academic standards, misconduct and regulation, authentic assessment at scale, and finishing up with the policy and practice that enables human learning.

So for those who couldn’t make it or just wanted to soak it up rather than scribble notes all day, here are some of the things we took away.

We’ve lost learning

Starting off the day Jim presented the findings of our research on students, assessment and AI. The top line was that the sector’s response shouldn’t start with what to do about AI – it needs to start with what assessment is actually for.

Our survey and focus groups revealed that students aren’t confused about AI – they’re making clever, informed and rational choices about how to work a system that has removed their incentives to learn. Instead of treading on old ground – do students use AI, what tools do they use, why do they use them – we set out to answer whether students understand what they’re learning.

47 per cent worry that their grades don’t reflect what they actually know, 38 per cent admit they sometimes submit work they couldn’t fully understand without going back to their sources, and 14 per cent said they’d feel anxious or worse if asked to give the reasoning behind their last submission without notes.

The extent to which students learn something through the assessment process rather than the output depends on how much of a role AI played in submission. Assessment design often means a student can turn on autopilot, coast through an assignment, learn nothing and move onto the next one, on the proviso they’re not re-assessed on that work later down the line.

And when it comes to accountability, it doesn’t stop students using AI – it changes how they use it. Any programme that has moved away from exams but hasn’t replaced the accountability function has removed the incentive to use AI effectively, along with the incentive to not use it at all.

When we asked what sort of assessment would work better for them, it was often one that tests learning and remains engaging – they didn’t want high-stakes exams. They instead wanted a whole redesign, with multiple attempts, uncapped formative practicum, peer explanation, and the confidence that opportunities – whatever they might look like – to show off learning is truly an opportunity, not a trap.

While a great deal of the sector still considers students to be cheating every time they open ChatGPT, there’s a real need to recognise what they’re actually doing is exposing the flaws in existing assessment design and filling in the gaps of institutional support and resources. Cheating, if that’s what you want to call it, is the new coping.

CD1

Prepared for what?

Universities are measured on their ability to produce “good” graduate outcomes. How do they prepare students with the skills and knowledge to enter the workforce?

Our research told us of students doing a year in industry or on placement being trained on AI tools, but then returning to the classroom to find those very skills are banned. If access to knowledge is being flattened, what skills do graduates actually value? What does it mean to be effective, employable and adaptable, and how is higher education preparing students for these realities?

One of our panellists started us off with the friendly reminder that AI probably won’t take your job but someone who can use AI might. Another outlined how students are rapidly building “shadow AI universities” – parallel worlds of learning practices in WhatsApp and Discord groups creating a two-tiered system across institutions.

To prepare graduates with AI skills, staff need to be confident in those skills too – often being trained in parallel with students due to the rapidly evolving nature of the technology.

But even though AI presents itself as accessible and universal, many universities are yet to extend premium forms of AI tools under licence to all staff and students. Those with the financial capital to buy premium software are at an advantage, and it’s not just financial capital either – for those who aren’t working more than 20 hours a week in part-time jobs, they’ll have more time to understand, experiment with, and develop these tools.

AI has the ability to be a leveller, but only when there’s sufficient support around it – otherwise it risks widening existing digital gaps.

CD2

Regulating the unregulated

One of the central moral panics about AI and assessment is academic misconduct. The Office of the Independent Adjudicator (OIA) published casework notes last summer showing that student complaints succeed when providers fail to fully explain AI misconduct conclusions or deny fair response opportunities.

And when universities properly implement students’ procedural rights – like actually sharing evidence in advance, allowing substantive responses, or providing detailed reasoning – it becomes even harder to prove and punish AI misconduct.

The burden of proof is typically on the university to prove the student has used AI, not for the student to prove they haven’t.

On AI detection software, most universities have now turned off the Turnitin AI checker, but other software flags international students, non-native English speakers, and disabled students who use legitimate language support or assistive technologies – which in the case of disabled students would be a reasonable adjustment.

In our session on academic standards, misconduct and regulation we explored whether current regulatory frameworks and complaints procedures are keeping pace with practice.

Jim presented representatives from the OIA and the Office for Students (OfS) with hypothetical case studies. Take Amara, an international postgrad student with dyslexia. Her Disabled Students’ Allowance (DSA) funds assistive software and a literacy support tool that suggests rewording sentences. Her dissertation receives an AI detection score of 67 per cent and she’s invited to a viva four months after submission.

She’s not shown any of the detection report and is asked to recall her process – including sources that she used months ago – which she struggles to do. The dissertation contained two references that couldn’t be located, which the university considered a “hallucination” and therefore AI use. Amara says she forgot to document the source but can do so now. She later provides planning notes, drafts, evidence of a previous misconduct case that incorrectly flagged her work, the role of her assistive technology, and the correct sources.

The panel upholds the allegation but their written reasoning doesn’t engage with Amara’s evidence or explain why it was rejected. The penalty is a mark of zero with capped resit, which isn’t explained against the range of available sanctions.

She appeals stating the university didn’t share evidence in advance, the viva was poorly designed, they didn’t consider her evidence, explain the reasons for the penalty, or consider detection tools for students writing with assistive technology or writing in a second language. The university rejects the appeal.

There’s a whole host of questions that follow around how this matches up with the OIA’s good practice framework and to what extent there are implications for B2 – resources and support. These hypothetical case studies are likely to ring bells for officers and advice teams who are regularly sitting in on misconduct panels.

CD3

Off script

Throughout the day we featured a number of student stories to centre the humans behind all the policy decisions institutions make. We heard from Bethany Jackson from Bucks SU, exploring the various reasons why students use AI – that it’s becoming a coping mechanism for students, not a cheating tool.

We then heard from Gemma Vael from Exeter Guild who identified the anxieties of creative students when it came to AI, the confusing university guidance and policies, and the importance of digital skills and literacy for when students enter the workplace. Gemma reminded the audience that students’ use of AI will never be homogenous.

After lunch we heard from Jeena Thomas at Anglia Ruskin SU who spoke about the international student perspective when it comes to academic misconduct processes. Their entire student journey is thrown in flux as they wait endlessly for an update, all while they listen to the tick-tock of their visa clock.

Finally Lee-Ann Durrant from Suffolk SU spoke about how she was allowed to use AI as a reasonable adjustment and how ChatGPT became the most useful reasonable adjustment that they were never offered.

Of course facts and stats matter and it’s easy to focus on these when it comes to something as complicated as AI, but it’s stories that bring feeling to power and can get people to better understand and pay attention.

CD4

Being your authentic self

Underpinning all the conversations about AI is learning and assessment. And AI has prompted many to revisit authentic assessment – to challenge students to think deeply about the outputs that evidence the things that matter, and to better prepare them for work beyond university.

The problem has always been that authentic assessment is hard to scale, so the question pivots to one of how do we scale oral exams, portfolios and real-world projects en masse?

At Aston University they build “power skills” into every undergraduate degree – which includes mastering AI and the latest digital tools, entrepreneurial thinking, being an inclusive leader, and learning to create a more sustainable planet. By integrating these skills across programmes students embrace AI by learning how to use it and building it into assessment.

Jayne Pearson from King’s College London (KCL) argues that AI hasn’t actually broken assessment but has exposed what it’s already ignoring. In her research she found that when students and staff were asked what the purpose of being assessed through essays was, they struggled to answer. And in a bid to tackle anxieties about the “writing” part of assessments being outsourced, institutions could embrace “processfolios” – which ask students to depict their journey of producing a piece of work through a collection of artefacts, such as drafts, plans, feedback, source notes, and AI prompts, accompanied by a reflection.

Jayne argues that what’s required instead is a cultural shift that aligns assessment, teaching and institutional messaging around writing as a process of learning, not just a product to be judged.