A man in a hi-vis vest and VR headset holds controllers in a training room.

Proving VR Safety Training Works: From Completion Records to Competency Evidence

Several OSHA standards already require proof that workers understood their training. Scored simulations, on-the-job checks and periodic re-tests can turn VR session data into a record that holds up to an auditor.

A training record says a worker attended. It does not say the worker can do the job safely. Most safety managers know this gap well, and most learning systems still report the same few things: who enrolled, who finished and who passed a short quiz.

Virtual reality (VR) training makes the gap harder to ignore. A headset session lets a worker isolate a pump, respond to a gas alarm or walk an evacuation route without real-world risk. It also records far more than a sign-in sheet ever could: which steps were taken, in what order, how long each took and where the worker hesitated. The question for safety leaders is no longer whether VR can deliver training. It is whether the evidence it produces shows competence, and whether that evidence would stand up when an inspector, an auditor or an investigator asks how you know a worker was ready.

This article sets out what U.S. employers are expected to show, what the research does and does not support, and a practical way to turn VR sessions into competency evidence rather than another completion log.

What OSHA already expects

OSHA standards rarely mention VR, and they do not need to. Several of them already ask employers to show that training worked, not just that it happened.

The Process Safety Management standard is the clearest case. Under 29 CFR 1910.119(g)(3), the employer must ascertain that each employee operating a covered process has "received and understood" the training. The training record must include the employee's identity, the date of training "and the means used to verify that the employee understood the training." Refresher training is required at least every three years, and more often if necessary (OSHA, 1910.119).

The powered industrial truck standard goes further. Under 1910.178(l), operators must be shown to be competent through training and an evaluation of their performance in the workplace. That evaluation must be repeated at least every three years. Refresher training is triggered by events such as unsafe operation, a near miss or a failed evaluation (OSHA, 1910.178). The standard lists interactive computer learning as acceptable formal instruction. It still requires practical exercises and a workplace evaluation alongside it.

OSHA's 2010 Training Standards Policy Statement adds a broader principle: required training "must be presented in a manner that employees can understand," in a language and vocabulary they can follow. It also tells compliance officers to check that this was done (OSHA, 2010). Hazard communication training under 1910.1200(h) must likewise be "effective" (OSHA, 1910.1200).

Two points follow for anyone adopting VR. First, a completion flag from a headset is not "the means used to verify" understanding unless the session actually tested something. Second, where a standard calls for a workplace evaluation, as 1910.178 does, a simulator score supports that evaluation but does not replace it.

What the research supports, and where it stops

The evidence for VR safety training is encouraging but narrower than marketing suggests.

A 2024 meta-analysis in the Journal of Safety Research compared VR with traditional construction safety training. It found VR significantly more effective, with standardized mean differences of 0.593 for behaviors, 0.432 for skills and 0.777 for experience measures. Training context and trainees' average work experience moderated the effect. The authors suggest VR for general safety training of young, inexperienced workers (Man, Wen and So, 2024).

A second review in Safety Science covered 52 studies across construction, firefighting, aviation and mining. It found VR outperformed traditional training for knowledge acquisition and retention. It also flagged weaknesses: only 9.6% of studies were explicitly grounded in theory, and only 36% reported long-term retention data (Scorgie et al., 2024).

The fair reading is this. VR tends to help workers learn and remember, especially newer workers. The research base says much less about whether that learning shows up months later on the plant floor. That gap is exactly what an employer's own evidence has to close.

What to measure: four layers of evidence

OSHA's own training guidance describes three types of evaluation: reaction assessments, learning assessments and training impact assessments (OSHA 3824). For VR, it helps to split learning into two parts, giving four layers. Each one answers a different question, and a completion record answers none of them.

Layer Question it answers Evidence to capture
1. Knowledge Does the worker know the hazards and the procedure? Scored checks on hazard recognition and procedure steps, in the worker's own language
2. Performance in simulation Can the worker carry out the task correctly without prompts? Steps completed in the right order, critical errors (for example, an isolation point missed), hints used, time to respond
3. Transfer to the job Does the worker do it the same way on the real equipment? A qualified observer's sign-off on the job, against the same checklist used in the simulation
4. Retention and impact Does it still hold months later, and is anything changing? Unassisted re-run at a set interval; trends in near misses, deviations and observations for the trained group

The most useful design choice is to use one checklist across layers 2 and 3. If the simulation scores the same critical steps that a supervisor checks at the pump or the truck, the two results can be compared directly. Where they disagree, the training or the simulation needs attention.

The second choice is to separate critical errors from everything else. A worker who takes 40 seconds longer than average is not the same as one who skips a lockout step. Pass rules should make critical errors disqualifying, however good the rest of the score is.

Building a record that stands up

A defensible competency record does not need new technology. It needs the right decisions to be made before the first headset goes on.

  1. Start from the task, not the scenario library. Write down the critical steps of the real procedure and the errors that would cause harm. Build or choose the VR exercise to test those steps.
  2. Set pass criteria in advance. Define what counts as competent, including which errors fail the attempt outright. Criteria set after the results are seen will not persuade an auditor.
  3. Run the session without hints for assessment. Practice sessions can coach. The assessed attempt should not, or the score measures the prompts rather than the worker.
  4. Check understanding in a language the worker uses. This follows directly from OSHA's 2010 policy statement. Narration, labels and questions in the worker's language matter as much as the graphics.
  5. Confirm on the job. Where a standard requires a workplace evaluation, VR prepares the worker for it and does not replace it. Even where none is required, a short observed task is the strongest evidence of transfer.
  6. Keep what the standard asks for. Record who was trained, when, by whom and how understanding was verified. Keep the scored attempt and the observer's sign-off with it.
  7. Re-test and review. Repeat an unassisted run before refresher deadlines, after a near miss or after a process change. Review the results with the people who do the work.

Two pitfalls are common. The first is treating time in headset as a measure of learning. It measures exposure. The second is letting the simulation drift from the plant. A scenario built on an old layout or a superseded procedure produces confident workers who have practiced the wrong thing. Any management-of-change review should include the training scenarios that depict the changed equipment.

From records to evidence

VR has made it practical to watch every trainee perform a hazardous task, something no classroom could offer. That is its real value for safety managers. Used only to generate completion certificates, it adds cost without adding assurance. Used to capture scored performance against a defined procedure, confirmed on the job and re-tested over time, it gives employers a far better answer to the question OSHA already asks: how do you know your workers understood?

References

Featured

Artificial Intelligence