How does AI read your "personality" through your conversations?

How does AI read your "personality" through your conversations?

07 May 2026, 08:51
5 min read
How does AI read your "personality" through your conversations?

A new study shows that AI is able to predict some of a user's core personality traits, such as a tendency to cooperate and emotional stability, with up to 60 percent accuracy, based on their conversation history.

In a recently published study, researchers from the  U.S. Federal Institute  of Technology Zurich (ETH Zurich) asked 668 ChatGPT users from the U.S. and the U.K. to share copies of their conversation logs, and trained an AI model to infer personality traits from these conversations.

The researchers collected and analyzed more than 62,000 conversations, categorizing them according to the topics they focused on.

The AI model, which was then carefully tuned, tried to estimate the likelihood that each user would have five personality traits known in psychology research as the "Big Five Traits": conscientiousness, conscientiousness, emotional stability, extroversion, and openness to experience.

Personality traits

Participants also underwent a standardized psychological test to determine their key personality traits.

 

The report showed that the recalibrated AI model was able to detect user personality traits with 61 percent accuracy; it was particularly adept at predicting the degree of compatibility and emotional stability, but had difficulty assessing conscientiousness.

The study showed that AI performs better when it has longer conversation logs to analyze, suggesting that more people use these models to recognise their personalities.

Risks of Use

Although the risks to individuals appear to be relatively limited, the researchers note that there is a "large-scale risk" if malicious actors exploit personality data.

The research team warns that this data could be used for "large-scale manipulation campaigns to spread disinformation or political propaganda."

The researchers hope their findings will contribute to the development of tools that reduce the risk of excessive personal data being shared with AI systems, such as systems capable of automatically removing identifying details.

Write a Comment

0 / 600

Comments (0)

Review Ranking →
No comments yet. Be the first to comment.