Work sample or situational judgment test: which should you use?
Use a situational judgment test when you need a cheap, standard screen for many applicants and you can supervise how they answer. Use a work sample when the job is knowledge work, a wrong hire is expensive, and you want to see how the candidate handles real material, including AI tools.
What each one is
A situational judgment test (SJT) shows a short workplace scenario and asks the candidate to pick or rank the best response from a list. It measures whether someone recognises a good response.
A work sample gives the candidate a real piece of the job: the documents, the data, the deadline. They produce the output themselves, a memo, a fix, a recommendation. It measures whether they can do the work.
| Situational judgment test | Work sample | |
|---|---|---|
| What the candidate does | Picks or ranks answers from a list | Produces real work from real material |
| What it shows | Recognising the right response | Doing the work, step by step |
| Cost to build | Low; off-the-shelf tests exist | Traditionally high; each role needs its own task |
| Cost to grade | Automatic | Traditionally slow; a trained reviewer per candidate |
| Works for hundreds of applicants | Yes | Traditionally no |
| Holds up when candidates use AI | Poorly; chatbots answer most items well | Well, if the task is built for AI use |
| How it feels to the candidate | Quick, can feel abstract | Longer, feels like the job |
What the research says
Both methods predict job performance. A 2022 re-analysis of decades of selection research, by Sackett, Zhang, Berry and Lievens in the Journal of Applied Psychology, placed work sample tests among the five strongest predictors of job performance, ahead of general cognitive ability tests.
The problem with SJTs today is not the research. It is the format. A multiple-choice scenario is exactly what language models are good at. In a 2025 study in Communications Psychology, six AI chatbots answered five emotional intelligence tests built in situational judgment format with an average accuracy of 81%, against a human average of 56%. If candidates take the test at home, you may be measuring their chatbot.
Why work samples were rare, and what changed
Selection experts have recommended work samples for decades, and most HR teams skip them. Building a realistic task for every role takes days. Grading open-ended work needs a trained reviewer for every candidate. With 200 applicants, nobody does that.
Two things changed. AI can now build a realistic task from a job description and grade open-ended work against a fixed rubric, so the cost has dropped. And AI tools made the old screens unreliable, so the cost of not seeing real work has gone up.
How WorkProbe does it
WorkProbe is a work-sample assessment. Paste in a job description and WorkProbe generates a 45-minute job simulation from it. Or bring your own scenario. The candidate gets a real task from the role, messy documents, simulated colleagues, and an AI assistant they are free to use.
We then grade what they did against a fixed rubric on four skills: role skills, AI collaboration, critical verification, and judgment under uncertainty. Grades are Strong, Acceptable, Weak or Poor. No scores, no percentages. You get a ranked shortlist and, for each candidate, interview questions anchored to what they actually did.
Allowing the AI assistant is the point, not a loophole. The simulation contains places where the assistant's answer is wrong or incomplete. You see who checks it, and who passes it on to the boss.
Which one to choose
- Choose a situational judgment test for high-volume frontline hiring, where you need one standard screen for thousands of people and you can run it supervised.
- Choose a work sample for analyst, operations, engineering, consulting and other knowledge-work roles; for senior hires where one wrong hire costs months; and whenever you cannot stop candidates from using AI.
- Use both if your volume is high: a basic screen first, a work sample for the shortlist that remains, then interviews.