AI Models Offer Conflicting Views on Job Vulnerability
Recent research indicates that advanced artificial intelligence models, including OpenAI's ChatGPT-5, Google DeepMind's Gemini 2.5, and Anthropic's Claude 4.5, provide inconsistent assessments regarding which professions are most susceptible to AI automation. This divergence in opinion has prompted economists to advise caution against blindly trusting AI-generated job-risk rankings.
Significant Disagreement on Key Roles
The study, conducted by economists Michelle Yin, Hoa Vu, and Claudia Persico, highlighted considerable discrepancies in how these leading AI chatbots evaluate job exposure. Notably, supervisory and management-type positions, as well as roles requiring a blend of mental and physical tasks, frequently elicited differing vulnerability scores.
For instance, while Anthropic's Claude suggested that accountants face high vulnerability to AI disruption, Google's Gemini strongly disagreed, rating the profession as less exposed. Similar disagreements emerged for roles like advertising managers and chief executive officers (CEOs).
Why AI Models Disagree on Job Exposure
Researchers attribute these inconsistencies primarily to the diverse datasets each AI model is trained on. Different training data can lead to varied interpretations of job functions and their inherent susceptibility to automation. The study also noted that current levels of AI adoption within certain professions might influence future exposure scores. For example, financial analysts already heavily utilize AI, which could lead future AI systems to perceive these jobs as more connected to and therefore more exposed to automation.
Expert Warning: Do Not Blindly Trust AI Assessments
Given these unreliable and inconsistent results, economists involved in the study strongly caution employers and individuals against relying solely on AI-generated job-risk rankings. Michelle Yin emphasized this point, stating, "I personally would not rely on just one measure to say, 'Oh, I should change my job,' or 'I should change my kid’s major.'" The findings underscore the complexity of evaluating AI's impact on the workforce and the need for human oversight and critical judgment.