What does the research actually say about whether immersive training works?
The honest claim is not that VR beats classroom teaching. It is that a well-matched simulation of a task you cannot safely or affordably rehearse is worth building.
What the pooled evidence shows
A 2026 second-order meta-analysis covering 26 meta-analyses, 180 studies and 18,792 participants reported an overall standardised effect of 0.55, dropping to 0.35 and non-significant in studies at low risk of bias, with modest publication bias and substantial unexplained heterogeneity.
A meta-analysis of learning from 360 degree video across 26 studies found no significant benefit nor detriment compared with other media.
What does predict a result
Task-technology fit, and designs that make the learner decide, retrieve or explain something. Display hardware has repeatedly failed to come out as a significant moderator, which is why the device is a deployment choice rather than the mechanism.
How to read our own numbers
Customer figures quoted on this site, such as the 94% first-time pass rate at Cardiff and Vale UHB, are real results from single programmes without control groups. They are service evaluations, not trials, and should be read that way.