The study
Who: KPMG and the University of Melbourne.
What: Trust, attitudes and use of AI: A global study 2025.
Type: A survey. People reported their own experience.
People: More than 48,000 people in 47 countries, surveyed from November 2024 to January 2025. The findings below come from the employees in the study who use AI at work.
What the researchers found
42% of employees who use AI at work said they had relied on AI output without evaluating its accuracy sometimes or more often.
The country reports also show how many employees said they had done this at least once. "At least once" is a lower bar than "sometimes or more often", so these figures are higher.
| Country | Relied on AI output without evaluating it, at least once |
|---|---|
| India | 81% |
| China | 78% |
| Singapore | 69% |
| Korea | 67% |
| Australia | 57% |
| Japan | 56% |
| New Zealand | 51% |
What this may mean for managers
This is our interpretation, not the researchers' finding.
Getting an answer from AI does not show that someone checked it. Checking every answer is not practical either. Teams need to agree which answers need a check, who does it and how much time it takes.
What this study did not test
- It did not test whether the unchecked answers were wrong. Many may have been right.
- It records what people reported. It did not observe their work.
Read next
- Questions to ask before acting on an AI answer
- Why an AI answer still needs a human decision
- Try a simulation where you decide what to check
Sources
- KPMG and the University of Melbourne (2025). Trust, attitudes and use of AI: A global study 2025. Global report. Country insights. Press release. Source checked 4 October 2026.