In short
Three questions on how far a person trusts a particular AI system. A 12-question long form exists. Use it for one named tool, and remember that more trust is not always better.
For whoever runs the assessment: an HR or operations lead, an internal project team, or an outside consultant.
What it measures
How much a person trusts one AI system that you name: the assistant in your support desk, say, or a forecasting tool.
Trust and attitude are different. A person can like AI in general and distrust the tool on their desk, with good reason.
How to use it
- Name the system in the survey introduction.
- Ask the people who use it.
- Compare the trust score with how often the system is right. High trust in a weak tool is a risk. Low trust in a good tool is waste.
- Look at the result by task. A person plus AI does not always beat either alone.
What it does not measure
- Whether the trust is deserved. A high score may mean the tool is good. It may mean people have stopped checking it. A review of research on trust in AI argues that the goal is trust that matches how reliable the system is, not maximum trust.
- Attitude to AI in general.
- Skill. People who understand how AI works choose better what to hand over to it.
How it was tested
The 12-question scale has been used for automated systems since 2000. In 2025 a research team in Australia tested it for several kinds of AI and built a three-question short form. Across four studies the short form agreed with other measures of trust and showed predictive value.
Licence and access
The 2025 article is open access under CC BY 4.0 and contains both forms. No other language version is established.
Alternatives
Other assessments with a similar job, and when to prefer each.
- How positive are staff about AI in general? AI Attitude Scale (AIAS-4) Prefer it when you want attitude to AI in general, not trust in one system. Staff, each about their own views · 4 questions Free to use and adapt
- What positive and negative views of AI do staff hold? General Attitudes towards Artificial Intelligence Scale (GAAIS) Prefer it when you want a fuller picture of positive and negative views. Staff, each about their own views · 20 questions, or 10 in the short form Free to use, ask before changing
The details
- Testing
- The long form is a long-established scale. One team re-tested it for AI and built the short form across four studies, and the results held up. No independent team has yet tested the short form.
- Looks at
- Person
- How the answers are collected
- A survey answered by each person about one named AI system.
- Length
- 3 questions; long form 12 questions
- Cost
- Free
- Licence
- CC BY 4.0You may translate and adapt the questions in the 2025 article if you credit the authors.
- Languages
- English
- By
- McGrath and colleagues, 2025
Sources
- McGrath, Lack, Tisch and Duenser (2025). Measuring trust in artificial intelligence: validation of an established scale and its short form. Frontiers in Artificial Intelligence. doi:10.3389/frai.2025.1582880
- Lockey, Gillespie, Morrill and Pool (2024). Trust in Artificial Intelligence: A Critical Systematic Review. Academy of Management Proceedings. doi:10.5465/amproc.2024.14990abstract
Related studies on this site
- Knowing how AI works improves decisions about what to hand over Pinski, Adam and Benlian (2023). AI Knowledge: Improving AI Delegation through Human Enablement. Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems.
- When a human plus AI beats either alone: 106 experiments Vaccaro, Almaatouq and Malone (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour, 8(12), 2293–2303.