Employees who use ChatGPT and other artificial intelligence tools at work may face a hidden career penalty, with managers less likely to trust them, according to new research. The findings suggest AI adoption can influence not only how work gets completed but also how supervisors assess employees’ credibility, performance and promotion potential.
The study, titled Psychological Reactions to Subordinates’ LLM Use at Work, was conducted by Samuel Meyers of the Hebrew University of Jerusalem, Yochanan E. Bigman of the same university, and Roman Briker of WHU – Otto Beisheim School of Management in Germany.
Published in The Leadership Quarterly, the research examined how employees’ use of large language models (LLMs), including ChatGPT, affects supervisory trust. The researchers conducted six studies, comprising five incentivised economic experiments and one field study involving academic supervisors.
AI use changes how managers assess employees
One experiment examined whether supervisors would accept an employee’s account of the time spent completing a task or independently verify the information.
Participants were told whether the employee had completed the work independently or used AI-generated output. The findings showed a significant difference in how willing supervisors were to trust the employee.
-
Employees who completed the task independently received supervisors’ trust 68% of the time.
-
When employees used AI, the likelihood of receiving trust fell to 24%.
-
AI-assisted work was still viewed more favourably than work copied from another person.
The experiment focused on whether supervisors trusted employees’ accounts, rather than assessing the quality of the finished work. This suggests the knowledge that AI was involved influenced perceptions of employee trustworthiness.
The researchers also identified consequences beyond immediate workplace trust. Across the studies, lower supervisory trust was associated with poorer perceived performance, lower assessments of organisational citizenship behaviour and reduced promotability.
In other words, employees’ use of AI could influence how managers perceive their contribution to the organisation and their suitability for career progression.
Clear AI policies can reduce the trust penalty
The research found organisational policy to be a significant factor in shaping managers’ reactions to employees using AI.
When organisations explicitly encouraged employees to use LLMs, the negative effect on supervisory trust was no longer statistically significant. However, the penalty remained when AI use was prohibited or when the organisation had not stated a policy.
The researchers also examined whether the importance of a task or a supervisor’s own experience with AI changed the outcome. Neither consistently reduced the trust penalty.
The findings indicate that workplace rules and organisational signals around AI adoption can influence how managers interpret its use, rather than technology familiarity alone.
Academic supervisors also showed lower trust
The final study extended the research beyond experimental settings. It involved academic supervisors evaluating work produced by graduate students.
Researchers found that the broader the students’ use of LLMs, the lower the supervisors’ trust in them. The result provided further evidence of the relationship between AI use and supervisory perceptions in an academic setting.
The findings raise questions for employers as AI tools become more common across business functions. While organisations may encourage employees to use these tools to improve efficiency, unclear policies could leave workers uncertain about how their managers view AI-assisted work.
Why employers need to address ‘shadow AI’
The researchers highlighted the possibility that employees may conceal their use of AI if they fear negative judgements from supervisors. Such undisclosed use is commonly referred to as shadow AI.
For organisations, the findings point to a need for more than simply making AI tools available. Clear policies on acceptable use and guidance for managers could help establish consistent expectations around AI-assisted work.
The study does not establish that every employee who uses ChatGPT will face lower trust or weaker promotion prospects. Its findings instead show how supervisory perceptions can shape workplace evaluations, and how explicit organisational encouragement can change this relationship.
As businesses expand their use of generative AI, how managers evaluate AI-assisted work may become an important part of workplace policy and talent management.
