AI models disagree on jobs most vulnerable to automation
What's the story
A recent study has revealed that leading artificial intelligence (AI) models, including OpenAI's ChatGPT-5, Google's Gemini 2.5, and Anthropic's Claude 4.5, don't agree on which jobs are most vulnerable to automation. The research was conducted by economists Michelle Yin and Hoa Vu of Northwestern University and Claudia Persico of American University. They analyzed how each model ranked professions based on their exposure to AI technology.
Divergence
Major disagreements between AI models
The study found major disagreements between the AI models, especially when it came to supervisory roles or jobs that combine cognitive and physical tasks. For instance, Claude rated accountants as highly vulnerable to AI automation, while Gemini gave the profession a much lower exposure ranking. The models also disagreed on the vulnerability of advertising managers and chief executives, highlighting their differing perspectives on job automation risks.
Impact
Future exposure scores influenced by AI adoption
The researchers also observed that AI adoption could influence future exposure scores. They suggested that occupations already heavily using AI, like financial analysis and digital office work, create more training data for newer models. This could potentially affect how these systems rate those professions in the future. The study cautioned against relying on a single AI model for high-stakes policy decisions around education, hiring, and workforce planning.
Caution
Caution against relying on single AI model
Yin cautioned against using a single AI-generated exposure score to make major career or education decisions. She said, "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.'" This highlights the need for a more nuanced understanding of how different AI models perceive job automation risks and their potential impact on various professions.