Can We Still Trust Hogan Assessments in the Age of AI?

Separating AI concerns from the evidence in talent assessment
A digital circuit board background featuring various icons related to technology and artificial intelligence. The icons include a shield with a checkmark, gears, a microchip, a person icon with a checkmark, and a brain with circuit patterns, all in blue and white against a dark backdrop. The central icon prominently displays "AI."

Author: Katja M. Friederichs

AI has changed the way we work, recruit, and assess talent. But it has also raised an uncomfortable question: if candidates can use an AI to answer assessment questions, can we still trust the results?

AI can solve reasoning problems, generate convincing answers, and even tailor personality assessment responses to specific job requirements. So, to answer that question, we need to distinguish between what AI is capable of and what the evidence tells us is happening.

Can AI fake a personality assessment?

Research suggests it can. In a 2024 study, researchers investigated whether large language models (LLMs), including ChatGPT, could manipulate personality assessment results to make applicants appear more desirable. When prompted to do so, ChatGPT often produced more favourable scores than human participants who were also trying to make a good impression.

Further a 2026 study found that two out of three AI models outscored people answering as job applicants across the personality assessment formats tested, including forced-choice formats designed to reduce response manipulation. Therefore, no assessment format should be assumed to be immune to AI-assisted cheating.

But demonstrating that AI can manipulate assessment scores is one thing. Understanding whether candidates are actually using it, and whether we can detect it, is another.

Can we tell when candidates use ChatGPT?

Hogan researchers put this to the test, asking ChatGPT to complete the Hogan Personality Inventory (HPI) as if it were applying for different roles.

ChatGPT showed considerably more consistent response patterns when prompted to apply for managerial roles than for professional or sales roles. This gave researchers a potential way to identify AI-generated responses. However, as AI models evolved, these promising detection methods quickly became outdated.

If detecting AI-assisted responses is becoming increasingly difficult, another question becomes even more important: is AI actually changing assessment results in practice?

What do more than 230,000 applicant assessments reveal?

In a study published in the peer-reviewed International Journal of Selection and Assessment, researchers analysed more than 230,000 Hogan assessments completed by job applicants, comparing results from just over two years before and after ChatGPT became publicly available in November 2022. The analysis covered personality (HPI) and two cognitive assessments: the Hogan Business Reasoning Inventory (HBRI) and the Hogan Judgement Assessment (HJA).

Research has shown that AI can achieve high scores on both cognitive and personality assessments, and ChatGPT has even outscored real job applicants on HPI scales. That makes the before-and-after comparison a meaningful test, because if candidates were using AI widely to improve their results, we would expect to see average scores increase.

However, the results showed no meaningful increase in assessment scores. While some changes were statistically detectable, they were negligible in practice and inconsistent in direction. Even among younger applicants, who might be expected to use AI more frequently, there was no meaningful pattern of score inflation.

In other words, while AI has the capability to influence assessment responses, the data currently provides little evidence of widespread AI-assisted cheating affecting Hogan assessment scores.

That is reassuring, but it doesn’t mean the risk can be ignored. As AI continues to evolve, protecting the integrity of assessment results remains essential.

How does Hogan approach this?

Hogan approaches this challenge through ongoing research, assessment design, and the interpretation of results.

  • Ongoing research and monitoring. Hogan investigates how AI tools respond to its assessments, explores ways to detect AI-generated answers, and monitors score trends across real applicant data as AI models evolve.
  • Assessment design that may discourage AI use. Hogan assessments incorporate design features, such as presenting questions one at a time, preventing respondents from revisiting previous answers, and imposing time limits on certain assessments. These features may make AI assistance more cumbersome.
  • Exploring results through feedback. Hogan assessments focus on reputation: how individuals are likely to be perceived by others at work. A feedback session allows practitioners to explore how well the results align with a candidate’s actual experiences and behaviour, potentially revealing inconsistencies that warrant further discussion. While this cannot reliably detect AI use, it adds a valuable layer of insight beyond the assessment scores.

Why honest answers benefit candidates too

A side of this debate that often gets overlooked is that candidates also have a stake in accurate results. A personality assessment isn’t an exam with right or wrong answers. It helps identify how well someone’s personality aligns with the demands of a particular role and working environment.

A candidate who uses AI to create a more desirable profile may appear better suited to a role on paper, while increasing the risk of ending up in a position that doesn’t match how they naturally work. That benefits neither the candidate nor the employer.

This is why Hogan has long emphasised the importance of answering honestly rather than trying to present an idealised version of themselves.

What does this mean for HR professionals?

Aggregate data cannot tell us whether individual candidates used AI, and cheating remains a possibility. For organisations using assessments in selection, a few principles are worth keeping in mind:

  • Separate possibility from prevalence. Just because AI can manipulate assessment responses doesn’t mean candidates are doing so at scale.
  • Consider proctoring when the stakes call for it. Where AI-assisted cheating is a serious concern, Hogan recommends administering assessments in a proctored setting.
  • Explain the purpose of the assessment. Help candidates understand that personality assessments are about finding the right fit, not achieving a perfect score. Being transparent about the process, including any subsequent feedback conversation, can encourage authentic responses.
  • Look at the bigger picture. Use feedback conversations and structured interviews to explore the results and treat Hogan assessments as one input alongside other relevant evidence rather than a standalone hiring decision.

The bottom line

The potential for AI-assisted cheating is real, and Hogan will keep monitoring assessment trends as AI tools become more capable and more widely used. But so far, the evidence is reassuring, suggesting that the concern is greater than the measurable impact.

 

Sources

Dabdoub, A., & Pool, R. (2026). Testing after ChatGPT: Aggregate change in applicant assessment scores. International Journal of Selection and Assessment, 34, e70077. doi.org/10.1111/ijsa.70077

Robie, C., Feeney, J. R., Rasheed, S., Bourdage, J. S., Christiansen, N. D., Dunlop, P. D., & Daljeet, K. N. (2026). Faking resistance of forced-choice and single-statement personality measures: Stress-testing with humans and generative AI. International Journal of Selection and Assessment, 34, e70075.  https://doi.org/10.1111/ijsa.70075

Phillips, J., & Robie, C. (2024). Hacking the perfect score on high-stakes personality assessments with generative AI. Personality and Individual Differences, 231, 112840. https://doi.org/10.1016/j.paid.2024.112840

Hogan Assessments (2025). ChatGPT and the Future of Assessments (technical brief).

Hogan Assessments (2025). Can job applicants fake personality tests? (blog).

Downloads:
Share article:
LinkedIn

You might also be interested in