For employability leads, careers service heads and curriculum leaders in higher and further education
An employability lead we spoke with recently put it plainly: students love the simulation sessions. Attendance is high, feedback forms are glowing, and the roleplay module is consistently the best-reviewed part of the course. What she couldn't answer, when asked by her Pro Vice-Chancellor, was whether any of it changed who got hired.
Feeling useful and changing outcomes are two different questions. This article is about the second one: what the evidence actually shows about simulation-based learning and employability, where that evidence is strong, where it's thin, and how an institution might measure its own impact rather than relying on satisfaction scores.
The pressure behind this question is not hypothetical. The World Economic Forum's Future of Jobs Report 2025 found employers expect 39% of core job skills to change by 2030, and 63% of them named the skills gap their biggest barrier to business transformation.
Meanwhile, 91% of learning and development professionals told LinkedIn's 2025 Workplace Learning Report that human skills matter more than ever, and 76% of employers told Cengage Group's 2025 Graduate Employability Report they're hiring the same number of entry-level staff as last year, or fewer. Institutions are being asked to prepare students for a skill set that keeps shifting, in a market that's tightening. Simulation is one answer. Whether it works is a separate question, and it's one the research can actually answer.
The strongest controlled evidence we found comes from a programme of trials on Virtual Reality Job Interview Training, known as VR-JIT, run by Matthew J. Smith and colleagues at Northwestern University. Participants rehearsed job interviews repeatedly against a virtual hiring manager and received structured feedback after each attempt, a close cousin of the deliberate-practice model behind most modern AI roleplay tools.
In a randomised controlled trial published in Psychiatric Services, 51% of trainees received a job offer within six months, compared with 25% of a comparison group who did not receive the training. After accounting for cognition, recency of employment and diagnosis, trainees were nearly ten times more likely to receive an offer (adjusted odds ratio 9.64, p<.05).
A related trial with people diagnosed with schizophrenia, published in Schizophrenia Research, found a similar pattern: 47.8% of trainees received a job offer within six months against 14.3% of controls. A larger 2024 dose-response study of 558 transition-age youth with disabilities found employment rose from 34.7% at low simulation dose to 61.3% at high dose.
We want to be direct about what this evidence does and doesn't show. These trials were conducted with people navigating vocational rehabilitation, including severe mental illness, schizophrenia and disability. That's a different population from a typical undergraduate cohort, and the difference matters for how the finding gets used.
We haven't found a randomised trial showing an equivalent employment-rate effect in a mainstream student population, and any provider who implies otherwise is overstating the evidence. What these trials do show is the mechanism: structured, repeated rehearsal with feedback measurably changes who gets hired, in the one setting where researchers have actually tracked people into employment rather than just asking how confident they felt afterwards.
“Trainees were more likely to receive a job offer than comparison participants after accounting for cognition, recency of last job and diagnosis.”
— Smith et al., Psychiatric Services, 2015
Randomised controlled trial comparing trainees who practised job interviews in a virtual simulation against a comparison group who did not.
The mechanism has three parts, and none of them are unique to VR or AI.
The first is deliberate practice: focused, repeated attempts at a specific skill with feedback between them, rather than passive exposure. A 2024 systematic review in Psychotherapy Research found deliberate practice consistently improved performance of discrete communication skills across the studies it examined.
The second is psychological safety. Live roleplay in front of peers carries real social risk, and that risk suppresses the number of attempts a nervous student is willing to make. Recent work on counselling skills training frames this explicitly: performance anxiety "may reduce participation, limit deliberate practice repetitions," and simulation design that removes an audience tends to increase how much practice actually happens.
The third is the experiential learning cycle Kolb described decades ago: concrete experience, reflection, conceptualisation, then a new attempt informed by what was learned. A 2025 systematic review of the model's use in higher education found that institutions frequently implement only part of the cycle, most often skipping structured reflection: precisely the step that turns an experience into a transferable lesson rather than just an interesting hour. It's the reason reflection sits alongside preparation and practice in how we structure a learning pathway, rather than treating the roleplay as the whole exercise.
There's an important caveat here, and it's one the research is unusually clear about. A 2024 review in the European Journal of Work and Organizational Psychology names it directly: soft skills training "often does not yield the desired behaviour changes at work," a pattern researchers call the transfer problem. Rehearsal builds capability. Whether that capability shows up in a real interview or on the job depends on how well the practice environment matches the real one, and on whether the skill gets reinforced afterwards rather than left in the simulator.
A 2024 study across two UK institutions, led by researchers at Edinburgh Napier University Business School and published in Studies in Higher Education, gives a useful answer to a question most simulation research skips: does it matter when in a degree students do this?
The researchers ran a team-based business simulation and measured self-assessed life skills before and after, building their measurement scale from interviews with graduate recruiters. The simulation had a strong, positive effect on first-year students' self-assessed life skills and subject knowledge. For fourth-year students, the effect was smaller and not statistically significant.
The likely explanation isn't that simulation stops working as students progress. It's that fourth-years have usually already had placements, part-time jobs and society leadership roles to draw skills from, so a simulation adds less on top of that. First-years, with comparatively little to draw on yet, gain the most from a safe, structured place to practise. The practical implication for curriculum design is straightforward: embed simulation early, and treat it as preparation for placements rather than a replacement for them later in the degree.
Most of the simulation literature focuses on what students learn. A newer strand looks at what they can prove.
If a student completes a structured simulation and their performance is scored against defined criteria, that performance becomes evidence a student can put in front of an employer, backed by a mark scheme rather than a completion certificate. This is the logic behind pairing simulation with micro-credentials or digital badges: the credential documents the assessment criteria and the specific skill demonstrated. In practice, that usually starts with a scored simulation drawn from a training library or built against an institution's own criteria, rather than an assessment framework designed from a blank page.
A 2024 study in Distance Education, based on 85 interviews with labour-market stakeholders, found employers generally view micro-credentials as a workable way to evidence soft skills, alongside real concerns about consistency and quality assurance across providers. A 2026 synthesis in Frontiers in Computer Science is blunter: the evidence base "is mostly cross-sectional and descriptive," and longitudinal proof that micro-credentials change hiring decisions is still thin. This is a layer worth building carefully, with realistic claims about what it currently proves.
The gap this leaves is well documented on the employer side too. NACE's Job Outlook 2025 and its analysis of new graduates' competency proficiency found students rate themselves far more proficient than employers do: 84.6% of students call themselves very or extremely proficient in professionalism, against 50% of employers, with gaps of a similar size for leadership, critical thinking and communication. That's the gap simulation is meant to close. The next question is how an institution would actually know whether it had.
Gap between how proficient students believe they are and how employers rate them, by skill area.
Most institutions we talk to are measuring the easiest thing to measure, which is how confident students felt afterwards. That's a legitimate data point, but it sits at the shallow end of what's known in training evaluation as the Kirkpatrick model: reaction, learning, behaviour, and results.
Reaction and learning, meaning whether students enjoyed it and whether their self-rated confidence went up, are worth collecting but easy to inflate and hard to defend to a Pro Vice-Chancellor asking about outcomes. Behaviour is the harder, more useful level: did an independent assessor watching a mock interview or observed roleplay actually rate the student's performance as better than before.
Results, meaning whether it changed employment, retention on a course, or progression, is the hardest to reach and the one most programmes never attempt, largely because it needs tracking over months, not a survey at the end of a session.
If your evaluation currently stops at students saying they felt more confident, treat that as a starting point. A finished business case needs the level below it: a sample of student performances scored against a rubric, blind to whether the assessor knows the student practised or not.
Being honest about limits is part of making a credible case for what simulation can do, so it's worth stating four of them plainly.
Simulation is not a substitute for a real placement. UK research linking work placements to graduate earnings, and a 2026 US matched-cohort study on paid internships and early-career outcomes, both find real workplace experience carries labour-market value that a simulated one doesn't fully replicate: an employer relationship, ongoing accountability, a result a student can point to on a CV.
Effects are weaker for students who already have workplace experience, as the Edinburgh Napier study above shows directly. A final-year student who has completed two internships needs different support than a first-year who hasn't yet had a formal interview.
Delivery matters as much as content. A body of employability research going back to Yorke and Knight's work on embedding argues that employability activities offered as optional extras serve the students who need them least, because the students who'd benefit most rarely opt in.
A 2022 study in Studies in Higher Education found embedding employability activity directly into a compulsory module improved career planning and confidence more than the same content offered as an optional bolt-on. A simulation sitting outside the curriculum, as something only keen students opt into, will reach the wrong students.
And AI roleplay specifically is still an early-stage technology in the research literature. A 2025 review in Interactive Learning Environments of 33 VR soft-skills studies found the field is short on control groups and long on self-report, particularly around how well a skill rehearsed with an AI character transfers to a real, high-pressure interaction. That's a reasonable basis for treating AI roleplay as one part of a wider design, paired with live practice and structured AI-driven feedback, rather than a stand-alone fix.
Simulation is still worth investing in. The pitch worth listening to, though, is the modest one. When you're evaluating a platform, ask what it can show you at each of the four measurement levels above, going beyond reaction and self-rated confidence.
Ask whether it's designed to sit inside a compulsory module or as an opt-in extra, given what the embedding research shows about who that choice reaches. Ask whether performance is scored against defined criteria a student could show an employer, or just logged as completed. And ask what the provider says its evidence can't yet prove: a provider willing to name the edges of the evidence is one more likely to have read it properly.
The same questions apply whether you're building this into a higher education programme or a further education one: the measurement bar shouldn't move because the qualification does. If you're building or reviewing a business case for simulation-based learning at your institution, we'd be glad to talk it through.
The best evidence shows that practising interviews in a simulation can help people get more job offers. In one study, people who practised with a virtual interview coach got a job offer 51% of the time, compared with 25% for people who didn't. This evidence comes from job-training programmes, not typical university students, so results at a university may look different.
Simulation-based learning means practising a real-life skill, like a job interview or a difficult conversation, in a safe, realistic setting before doing it for real. Students can try, get feedback, and try again without the risk of a real interview going wrong.
A normal role play uses another person to play a role, like a classmate or tutor. A business simulation is usually a longer group exercise, like running a pretend company. AI roleplay uses a computer program to play the other person, so a student can practise alone, as many times as they like, and get feedback straight away.
The most complete way is to check four things: did students like it, did they feel more confident, did an assessor watching them see a real improvement, and did it change their chances of getting a job or a placement. Most programmes only check the first two, which is not enough to prove it works.
1. Smith, M.J., Fleming, M.F., Wright, M.A., et al. (2015). "Job Offers to Individuals With Severe Mental Illness After Participation in Virtual Reality Job Interview Training." Psychiatric Services, 66(11), 1173–1179. https://psychiatryonline.org/doi/10.1176/appi.ps.201400504
2. Smith, M.J., Fleming, M.F., Wright, M.A., et al. (2015). "Virtual reality job interview training and 6-month employment outcomes for individuals with schizophrenia seeking employment." Schizophrenia Research, 166(1-3), 86–91. https://pmc.ncbi.nlm.nih.gov/articles/PMC4512907
3. Smith, M.J., Van Ryzin, M., Jordan, N., et al. (2024). "Virtual Job Interview Training: A Dose Response to Improve Employment for Transition-Age Youth With Disabilities." Career Development and Transition for Exceptional Individuals, 47(2), 92–105. https://pmc.ncbi.nlm.nih.gov/articles/PMC11232528
4. Thompson, J. et al. (2024). "Enhancing graduate employability – exploring the influence of experiential simulation learning on life skill development." Studies in Higher Education, 50(2), 256–270. https://www.tandfonline.com/doi/full/10.1080/03075079.2024.2334837
5. Sheen, J. et al. (2024). "The influence of deliberate practice on skill performance in therapeutic practice: A systematic review of early studies." Psychotherapy Research. https://www.tandfonline.com/doi/full/10.1080/10503307.2024.2308159
6. "The SAFE Fishbowl Model: Psychological Safety by Design in Counseling Skills Training." (2025). Journal of Counselor Preparation and Supervision, 20(3). https://research.library.kutztown.edu/jcps/vol20/iss3/8/
7. "Applying Kolb's experiential learning cycle for deep learning: A systematic literature review." (2025). Social Sciences & Humanities Open. https://www.sciencedirect.com/science/article/pii/S2590291125008253
8. Weick, M. et al. (2024). "Making soft skills 'stick': a systematic scoping review and integrated training transfer framework grounded in behavioural science." European Journal of Work and Organizational Psychology, 237–250. https://www.tandfonline.com/doi/full/10.1080/1359432X.2024.2376909
9. "Kirkpatrick Model: Its Limitations as Used in Higher Education Evaluation." ERIC EJ1291600. https://files.eric.ed.gov/fulltext/EJ1291600.pdf
10. Bruguera, C., Pagés, C. et al. (2024). "Micro-credentials and soft skills in online education: the employers' perspective." Distance Education, 56–76. https://www.tandfonline.com/doi/full/10.1080/01587919.2024.2435645
11. "Micro-credentials in higher education: transforming credentialing for lifelong learning and workforce alignment." (2026). Frontiers in Computer Science, 8. https://doi.org/10.3389/fcomp.2026.1853493
12. "The impact of work placements on graduate earnings." (2023). Studies in Higher Education. https://doi.org/10.1080/03075079.2023.2211999
13. "Work-Based Learning and Early Career Outcomes: Evidence on Earnings, College-Level Employment, and Career Progression." (2026). Research in Higher Education. https://link.springer.com/article/10.1007/s11162-026-09888-4
14. Yorke, M. & Knight, P. "Embedding Employability into the Curriculum." ESECT / Higher Education Academy. https://www.qualityresearchinternational.com/esecttools/esectpubs/Embedding%20employability%20into%20the%20curriculum.pdf
15. "Evaluating the efficacy of embedding employability into a second-year undergraduate module." (2022). Studies in Higher Education. https://doi.org/10.1080/03075079.2021.2020748
16. Kamińska, D., Zancanaro, M. et al. (2025). "Virtual reality for the training of soft skills for professional education: trends and opportunities." Interactive Learning Environments, 33, 3261–3281. https://www.tandfonline.com/doi/full/10.1080/10494820.2025.2450634
17. World Economic Forum (2025). Future of Jobs Report 2025. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
18. National Association of Colleges and Employers (2025). Job Outlook 2025 and "The Gap in Perceptions of New Grads' Competency Proficiency." https://www.naceweb.org/research/reports/job-outlook/2025
19. Cengage Group (2025). 2025 Graduate Employability Report. https://www.cengagegroup.com/news/press-releases/2025/cengage-group-2025-employability-report/
20. LinkedIn (2025). Workplace Learning Report 2025. https://business.linkedin.com/learn/resources/workplace-learning-report