Traditional, off-the-shelf, unsupervised psychometric tests are increasingly vulnerable to AI assistance. However, well-designed, role-specific assessments that combine psychometrics with verification can remain strong indicators of job performance.
1. Where AI helps — and where it struggles
More vulnerable to AI
- Verbal and numerical reasoning: standard, text-based questions can easily be copied into an AI tool for assistance.
- Written responses: AI can generate polished answers to questions about experience, stakeholder management or problem-solving that may sound credible without reflecting the candidate’s actual experience.
Less vulnerable to AI
- Abstract reasoning: visual and pattern-based tasks are harder to outsource, particularly when questions are dynamic.
- Personality and values assessments: candidates can try to present themselves favourably, but consistently manipulating responses across a well-designed questionnaire is difficult.
- Situational judgement tests: role-specific scenarios — such as handling a tenant complaint, prioritising competing campaigns or resolving an operational issue — require context, judgement and trade-offs that are harder for AI to replicate convincingly.
2. How assessment providers are responding
Good providers are increasingly designing assessments with AI risk in mind. This includes:
- Security controls: restricting copy/paste, right-clicking and tab switching, with some platforms also monitoring screenshots and unusual browser activity.
- Behavioural analytics: identifying unusual patterns such as extremely rapid responses, long pauses followed by large amounts of text, or inconsistent performance across similar questions.
- Adaptive testing: rotating questions and adjusting difficulty makes it harder to share answers or prepare specifically for a test.
- More realistic assessment: using case studies, scenarios and tasks that reflect the actual demands of the role.
- Proctoring – providers offer AI‑assisted proctoring that can take periodic photos or short video clips during the session and track IP addresses and flag suspicious location changes.
- Process-based evaluation: assessing how candidates approach a problem, rather than simply checking if they answered the question correctly.
These measures do not eliminate AI-assisted cheating, but they make assessments harder to game and provide additional signals about authenticity.
3. Why psychometrics still have value
AI has not made psychometric testing obsolete. The key issue is how the assessment is designed and how its results are used. Generic, static, unsupervised tests are more exposed to AI. In contrast, adaptive, role-specific and scenario-based assessments are harder to manipulate and more closely aligned to actual job requirements. Most importantly, psychometrics should be one signal, not the verdict. Combining them with structured interviews, work samples, references and other verification steps provides a much stronger basis for hiring decisions.
The goal is not to make AI impossible to use. It is to design assessments where genuine capability is difficult to fake and easy to verify.