Artificial intelligence can now predict how you’ll answer a personality questionnaire — even before it’s written. A new study reveals that ChatGPT, specifically GPT-4, can generate valid personality tests from any source text and accurately forecast population-level responses, suggesting that AI has absorbed an expert-level understanding of human psychology just from reading the internet.
The Research
Researchers from the Hebrew University of Jerusalem, led by Rotem Monsa, published findings in the Cell Press journal iScience on August 6, 2026. They used GPT-4 to create two personality questionnaires: one based on the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), the gold-standard clinical text for mental health, and another based on an astrology textbook as a control.
The DSM-5-derived questionnaire asked participants to rate statements like “often suspects others’ motives” on a scale from 1 (strongly disagree) to 5 (strongly agree). The astrology-based questionnaire similarly drew trait descriptions from zodiac signs. Despite the unscientific source, ChatGPT extracted trait-related language that still predicted real-world outcomes like depression, anxiety, and well-being at levels comparable to the standard Big Five Inventory (BFI).
Remarkably, GPT-4 predicted the average responses and the inter-item correlation matrix of 600 human participants before they took the surveys. The DSM-5-derived questionnaire showed high internal consistency, matching the BFI’s structure, while the astrology-derived one predictably lacked coherence—mirroring real-world psychometric validity.
Why It Matters
This breakthrough suggests that large language models (LLMs) have “learned the structure of human personality as a natural byproduct of their training,” as Monsa explains. These models aren’t explicitly taught psychology but infer it from language patterns in billions of web pages, social media posts, and books. This means AI could rapidly develop and validate surveys for psychological research, marketing, or even personalized brain training—without needing months of human pilot testing.
But there’s a caveat: The current predictions are most accurate for Western, English-speaking populations. The authors caution that performance may drop across other languages and cultures, a limitation being actively studied.
What You Can Do
Understanding personality is key to knowing your cognitive strengths and areas to improve. While AI can predict group trends, your individual profile is unique. Take a validated test, like the Big Five Inventory, and reflect on how your traits influence your daily thinking, learning, and decision-making. For deeper insight, consider a free cognitive assessment to see how your brain performs across multiple domains.
Source: Neuroscience News
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.
