How to Test AI Skills When Hiring: Prompting, Judgment and Responsible Use
Ismayil Ismayilov7 min read
Key takeaways
- Practical AI skill is not "knowing ChatGPT exists". It's writing good prompts, checking the output and using AI safely.
- Test three areas: prompting, critical evaluation of AI output and responsible use (privacy, security, ethics).
- Tailor the depth by role: office roles need productivity skills, analysts need data skills, and product or engineering roles need deeper knowledge.
- Keep AI skill tests separate from regular skill tests, where AI use counts as cheating.
Why test AI skills now
AI assistants are part of daily work in sales, marketing, finance, HR, support and engineering. Two candidates with the same CV can differ hugely: one saves hours a week with AI, while the other copies confident but wrong answers into client emails.
What to measure
| Area | What good looks like | Example question |
|---|---|---|
| Prompting | Clear goal, context, format and constraints | "Which prompt will produce a 5-line customer summary in Azerbaijani?" |
| Checking output | Spots wrong facts, invented sources, bad numbers | "The AI summary says revenue grew 40%. The table shows 14%. What do you do?" |
| Data privacy and security | Doesn't paste confidential data into public tools | "Which of these can you safely paste into a public chatbot?" |
| Choosing the right tool | Knows when AI helps and when it doesn't | "Which task should NOT be delegated to an AI assistant?" |
| Ethics | Avoids bias, discloses AI use where needed | "An AI shortlists candidates. What must a person check?" |
By role
| Role | What to emphasise | TheyTest library tests |
|---|---|---|
| Office, admin, HR | Productivity, privacy | AI Productivity for Office Roles; AI Tools Fundamentals (Workplace Use) |
| Sales, marketing, management | Business prompting | Prompting for Business Tasks |
| Analysts | Data interpretation with AI | AI for Data Understanding & Analysis |
| Everyone handling data | Safe use | AI Ethics, Privacy & Workplace Security |
| Product / engineering | Building with AI | AI Product Engineering Assessment; Prompt Engineering Fundamentals |
AI skills test vs AI cheating
In an AI skills test, the questions are about using AI well, and candidates still answer them themselves.
In a regular skills test (Excel, SQL, accounting), using AI to answer is cheating. Keep the rules clear for both:
- tell candidates up front that external tools aren't allowed during the assessment
- keep full-screen, tab-switch detection and session recording on for every test
A sample assessment for an office role (25 minutes)
- AI Productivity for Office Roles (12 min)
- AI Ethics, Privacy & Workplace Security (12 min)
- One short video question: "Describe a task where you used an AI tool at work. How did you check the result?"
FAQ
What is an AI skills test?
A test of how well a person uses AI tools at work: writing prompts, checking the output and protecting data.
Should every candidate take an AI skills test?
For most office and knowledge roles in 2026, a short one is useful. For roles that build AI products, use a deeper technical test.
Can candidates use ChatGPT during an AI skills test?
No. The test measures their own judgment about AI. Use the same anti-cheating settings as for any other test.
