University of Waterloo study suggests one type of warning can be effective
Are you worried about job candidates using generative AI in the application process?
Of course, most recruiters should be, given the massive rise in AI use by jobseekers and workers alike. The tool’s ability to mask negatives and boost positives should definitely be a concern when it comes to evaluating potential.
So, a study out of the University of Waterloo provides some comfort, in showing that a simple warning to job applicants can help curtail some use of AI.
Published in the International Journal of Selection and Assessment, the 2026 study tracked how often job seekers used generative AI tools such as ChatGPT during standardized hiring assessments. Not surprisingly, fewer than 3% of applicants admitted to using AI in 2024, compared to over 23% in 2025.
Three types of warnings for job candidates
But recruiters keen to know the true candidate and their skills and aptitudes will be encouraged to hear that a short warning placed before the assessment managed to reduce AI use to about 10%, according to the study.
What did the warning say? The researchers tried three different types to assess which worked the best. The first “explicitly” noted that AI use was prohibited and responses might be reviewed, so it was intended to appeal to a person’s motivation to avoid negative consequences.
The second type of warning was an “educational” one that emphasized every participant should have an equal opportunity to show their strengths, and that using genAI “undermines the integrity of the process by creating unfair advantages,” according to the researchers.
The third type tested was a “reasoning” warning that encouraged applicants to “respond authentically” and emphasized that honest responses would “best reflect their true potential and ensure a better match with the role.”
In the end, the researchers found that the consequence-based warning was the most successful, nearly halving the proportion of self-identified genAI users:
“To support a fair and transparent assessment process, we ask that you answer all questions independently and to the best of your ability. The use of outside assistance—including generative AI tools (such as ChatGPT or Gemini)—is strictly prohibited. Such actions undermine the integrity of the results and may lead to further review of your answers. We appreciate your understanding and support in following these guidelines.”
No backlash but guardrails needed
Interestingly, they also found that this "cost-effective tool to deter genAI use” doesn’t put off candidates from the hiring process, despite prior research indicating that highly punitive warnings can lead to negative applicant reactions.
“Warnings did not appear to diminish applicants’ perceptions of fairness or to generate backlash. This suggests that in applied assessment settings, consequences-based genAI warnings can be implemented without undermining applicant reactions — though careful attention to tone remains important,” say the researchers.
However, they also say that the warning primarily acts as a gatekeeper, reducing the number of job applicants who use genAI “without fundamentally altering the motivations of those who proceed despite the prohibition.” As a result, employers looking to deter AI use should implement clear, upfront messaging about the consequences but not assume this will eliminate the behaviour entirely, says the study.
“Given the escalating adoption of genAI tools, organizations may need to combine warnings with innovations in assessment design and monitoring to safeguard validity.”
Finding a balance to deter AI use
It’s also interesting to note that the researchers found applicants primarily turned to AI not to save time or reduce stress, but to handle cognitively demanding questions or clarify confusing instructions — suggesting that genAI use “may stem from genuine confusion or uncertainty about assessment demands."
As a result, they suggest employers should consider if the implementation of better instructions, practice items or enhanced accessibility features could also reduce the use of AI by job applicants.
Another interesting finding? The research found no meaningful link between AI use and higher job fit scores. The one exception: applicants who used AI specifically to cope with stress or anxiety tended to score lower, suggesting that stress-driven AI use "may reflect broader individual differences — such as lower confidence, weaker problem-solving skills or difficulty managing ambiguity."
It’s a useful insight but it’s unclear how recruiters would know why certain applicants were using the AI.
Overall, the researchers suggest that “overly aggressive detection or punitive enforcement strategies may disproportionately affect applicants who used GenAI for relatively benign purposes.”
I’d say the findings are encouraging in suggesting that many people will respond to warnings and act accordingly. But clearly this isn’t the only solution, or the best one, so HR might want to look at their assessment tools and messaging to ensure they’re staying on top of this massive wave of AI use.
