In recent years, artificial intelligence (AI) powered tools and chatbots have become more than personal assistants. The emergence has triggered fear among employees about AI replacing several jobs. Therefore, to get a deeper understanding of how AI could impact certain jobs, economists and researchers asked advanced AI models about jobs that are most vulnerable to AI. However, the results appeared to be unreliable. 

A new working paper, reported by The Wall Street Journal, was authored by economists Michelle Yin and Hoa Vu of Northwestern University and Claudia Persico from American University, who tested OpenAI’s ChatGPT-5, Google DeepMind’s Gemini 2.5, and Anthropic’s Claude 4.5 to see how each model ranked professions based on their exposure to AI.

Must read: OpenAI employees cash out billions as AI boom creates new tech millionaires

Accountants, CEOs and ad managers: Jobs AI models disagreed on

According to the report, different AI models gave very different opinions on which jobs are most at risk from AI automation. However, AI chatbots mostly disagreed on supervisory or management-type jobs and jobs that require both mental and physical work, on how vulnerable they are to AI.

Anthropic’s Claude stated that accountants could be highly vulnerable to AI disruption. Whereas Gemini highly disagreed, saying that it is less exposed to AI risk. It also disagreed on jobs like advertising managers and CEOs.

The study further highlighted that ChatGPT and Gemini mostly provided similar responses, but they still disagreed around 25% of the time. Researchers suggested that the difference could appear due to the different kinds of data they learn from. 

Must read: AI disruption, work pressure pushing BFSI employees towards exit, says report

For instance, professions like financial analysts use AI heavily, highlighting that future AI systems may assume those jobs are more connected to AI and therefore more exposed to automation. Therefore, greater AI adoption may also shape future exposure scores. 

Based on the study, economists warn that employers should not blindly trust AI-generated job-risk rankings, as the results can be inconsistent. Economist Michelle Yin 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.”