Ask a vendor whether their VR safety training makes you OSHA compliant and you will usually get an enthusiastic yes. That answer is wrong, and believing it is how EHS managers end up with a citation and a very expensive surprise.
The honest answer is: VR can satisfy part of what OSHA requires, and there is one part it structurally cannot touch. Knowing exactly where that line falls is the difference between a training program that survives an audit and one that quietly fails it.
We build industrial XR — we built a Work at Height simulator on Meta Quest 3 in ten weeks — so this is not a case against VR. It is a case against the sales pitch.
The short version
Most OSHA standards that require training break it into three parts:
- Formal instruction — lecture, discussion, video, written material, computer-based training.
- Practical training — demonstrations by the trainer, and exercises performed by the trainee.
- Evaluation — someone qualified watches the worker perform the task in your workplace and certifies they can do it.
VR is genuinely good at (1). It is often excellent at (2). It cannot do (3), because (3) is defined as happening in the actual workplace, on the actual equipment, observed by a person who takes responsibility for the judgment.
No headset can accept that liability.
Where this is written down
The clearest example is powered industrial trucks — forklifts — under 29 CFR 1910.178(l). It is worth reading because the structure repeats across many standards.
The rule requires that training consist of formal instruction, practical training, and "an evaluation of the operator's performance in the workplace." It then requires the employer to certify that each operator has been trained and evaluated, and that certification must include:
- the name of the operator
- the date of the training
- the date of the evaluation
- the identity of the person performing the training or evaluation
Notice what that last line does. It requires a named human being to attach themselves to the claim that this worker is competent. That is the part vendors gloss over.
OSHA has said in letters of interpretation that computer-based training can be part of formal instruction, but that it cannot be the only method — and that it does not substitute for the hands-on evaluation. VR is a more convincing form of computer-based training. It is not a different legal category.
The thing nobody advertises
There is also no such thing as an OSHA-approved VR course. OSHA does not approve, certify, endorse, or bless training products. If a vendor's marketing page says "OSHA approved," that alone tells you how carefully they read the standards.
What OSHA does is set requirements for content and documentation. The obligation to meet them stays with you, the employer. You cannot outsource a citation.
What VR is genuinely better at
None of the above makes VR a nice-to-have. For high-consequence, low-frequency tasks, it is the strongest tool available, for a specific reason: it lets people fail.
In a classroom, a worker who forgets to clip their harness loses a point on a quiz. In a properly built simulation, they fall. That stomach-drop moment does something a slide deck cannot — it converts a memorized rule into a physical reluctance to repeat the mistake.
VR is also the only practical way to train some scenarios at all. You are not going to stage a real arc flash, a real confined-space atmosphere failure, or a real fall from height so that 500 employees can each experience it once.
So the sensible position is not "VR or the floor." It is: use VR to make the formal and practical training dramatically better, and stop pretending it closes the evaluation.
The measurement problem underneath all of this
Here is the deeper issue, and it is not really about VR.
Almost every training platform — LMS, VR, classroom sign-in sheet — reports the same metric: completion. Did the person get to the end? Tick. Certificate issued. Dashboard turns green.
Completion is a record of attendance wearing the costume of a result.
An auditor asking "can this person do the job?" is not asking whether a module was finished. Neither is the incident investigator asking why a trained operator did the thing they were trained not to do. Both are asking about competency — demonstrated ability to perform the task correctly, including when conditions are bad and nobody is watching.
The gap between those two words is where most safety programs actually live:
- The certificate says trained. The audit finding says otherwise.
- The completion rate is 98%. The near-miss log disagrees.
- The expert who actually knew how it was done retires, and the know-how leaves with them.
If your training system only records completion, you have no evidence for the question anyone important will ask.
What an auditor will actually ask you for
In our experience the questions are boringly consistent, and none of them are about how realistic your graphics were:
- Who was trained, on what, and when? Records per person, per task — not a course roster.
- Who evaluated them, and when? A name and a date. For forklifts this is explicit; for most standards it is the practical expectation.
- What did the evaluation consist of? "They passed" is not an answer. What was observed, against what criteria?
- What happened after a near miss or a change in equipment? Refresher training triggers are written into many standards and are the most commonly missed requirement we see.
- Can you produce this in under an hour? Records that exist but cannot be retrieved during an inspection function, practically, as records that do not exist.
Run your current stack against that list. If any answer is "we'd have to pull that together," that is the actual gap — not the headset.
Buying advice, stated plainly
If you are evaluating VR safety training vendors, the demo will be designed to impress you with fidelity. Fidelity is the easiest thing to sell and the least predictive of outcomes. Ask these instead:
- What data comes out of a session, per trainee? Not a completion flag — the actual behavioral record. What did they do, in what order, and where did they hesitate?
- Does it map to the specific standard and equipment class we operate? OSHA requires training on each type of truck an operator will drive; a generic "forklift module" may not cover your fleet.
- How does the evaluation get recorded, and by whom? If the vendor has no answer, you are buying instruction, not compliance.
- What happens on failure? Does the system distinguish "got it right on the third try" from "got it right immediately"? Those are different workers.
- Can we export the record? If the evidence lives only inside their platform, your audit trail is a vendor relationship.
Being told where a product stops is more useful than being told what it does. A vendor comfortable telling you what they cannot do is generally the one worth talking to.
Where we sit on this
We build XR training, and we are also building LSM — which exists because of the gap described above. Traditional platforms record that someone watched a course. LSM is a browser-native way to measure competency: proof that a person can actually do the job, captured as behavioral evidence rather than a completion tick.
It is patent-pending and pre-launch, arriving in 2026 with a limited early-access cohort. It does not remove your obligation to have a qualified person evaluate operators in your facility — nothing does. What it is built to change is the quality of the evidence you bring to that evaluation, and what you can show an auditor afterwards.
If the "completion is not competency" problem is one you recognise from your own numbers, that is the thing worth a conversation — get in touch or look at LSM.
This article describes general requirements under US OSHA standards and is not legal advice. Standards are updated, and state plans may impose stricter requirements than federal OSHA. Verify the specific standard that applies to your operation, and consult your safety counsel or compliance officer before making program decisions.