🎗️Beyond Text and Images: ChatGPT Predicts The Results of Human Personality Tests

What if artificial intelligence could create a personality test, and even predict how people would respond before they even took the test? A new study put this idea to the test with 600 adults and found that artificial intelligence was able to identify human patterns in language and transform texts into questions capable of assessing personality traits.

Human personality seems difficult to measure: after all, how do you transform characteristics like extroversion, anxiety, organization, or openness to new experiences into questions that can be answered in a questionnaire?

A new study investigated a surprising possibility: using artificial intelligence to create personality tests from existing texts. The researchers wanted to find out if a language model would be able not only to write questions, but also to predict how people would respond to them.

To understand what the researchers did, you first need to know the so-called Big Five Factor model of personality. It organizes the human personality into five major characteristics:

1- openness to new experiences

2- responsibility and organization

3- extroversion

4- ease of coexistence

5- tendency to experience negative emotions

This model arose from the idea that important personality characteristics end up appearing in the very language we use to describe people. Therefore, the researchers asked: if language contains clues about personality, could an artificial intelligence trained on enormous amounts of texts also identify these clues?

The team then created a system in which artificial intelligence analyzed different text sources and, based on them, produced questions to assess personality traits. Two very different sources were chosen. One was the manual used by mental health professionals to describe personality disorders, which contains descriptions of psychological characteristics, the DSM-5.

The other was a popular astrology book. The choice was deliberate: the researchers wanted to compare a source based on scientific knowledge with another that has organized personality descriptions but lacks scientific foundation.

After creating the questions, the researchers conducted an important test: 600 adults answered the questionnaires. Each participant answered both the questions created from the descriptions in the mental health manual and the questions based on astrology. They also answered a traditional and widely studied questionnaire on the Big Five personality traits.

In this way, the scientists were able to compare the performance of the questions created by artificial intelligence with that of an already established psychological instrument. They analyzed, among other things, whether the questions in each questionnaire actually worked together and whether the answers could be related to real-life characteristics and outcomes.

But there was an even more interesting step. Before the 600 volunteers even answered the questionnaires, the artificial intelligence attempted to predict patterns in human responses. In other words, the researchers didn't just want to know if the artificial intelligence could write good questions. They wanted to find out if the model could already anticipate how people would tend to answer them.

The results indicated that the model was able to predict response patterns in both questionnaires created, showing that it had captured certain regularities present in human language.

However, the results do not mean that artificial intelligence has discovered a perfect new way to diagnose personality. The questionnaire based on descriptions from the mental health manual showed good consistency, similar to the traditional Big Five personality test. The questionnaire created from astrology, however, showed much less consistency.

Interestingly, some questions from both questionnaires managed to predict aspects of people's lives at levels comparable to those of the traditional questionnaire. This suggests that artificial intelligence can find useful information about personality hidden in language, but it also shows that the origin and organization of the information used to create the questions make a difference.

In the end, the study presents a rather interesting possibility for psychology: in the future, artificial intelligence could help researchers transform large volumes of text into new instruments for studying personality. But this doesn't mean that any questionnaire created by artificial intelligence is automatically reliable.

Before being used, it needs to be carefully tested with real people to verify if it truly measures what it promises to measure. The main finding, therefore, is not that artificial intelligence "knows" human personality, but that it can capture language patterns related to how people think, feel, and behave.

READ MORE:

Generating and analyzing personality questionnaires using large language models

Rotem Monsa, Aviv Zohar, and Shahar Arzy

iScience. Volume 29, Issue 8116909, August 21, 2026

DOI:10.1016/j.isci.2026.116909

Abstract: 

The Five-Factor Model (“Big Five”) is based on the lexical hypothesis that personality traits are encoded in language. Large language models (LLMs) offer new ways to explore personality through text. We developed an LLM-based method to generate and validate personality questionnaires from textual sources. Using the DSM-5 personality disorders section and a popular astrology book, we generated two questionnaires and administered them, alongside the Big-Five Inventory (BFI), to 600 adults. Internal consistency was high for the BFI and the DSM-based questionnaire but low for the astrology-based one. The LLM predicted human response patterns for both generated questionnaires. Individual items from both the DSM- and astrology-based questionnaires predicted diverse life outcomes at levels comparable to the BFI. These findings show that LLMs can construct personality measures and anticipate response patterns, offering a scalable framework for corpus-based personality research.