Can AI feel anxious? Study finds ChatGPT mirrors human stress patterns; Mindfulness techniques help it ‘calm down’

Artificial intelligence (AI) doesn’t have emotions, but a new study suggests it can exhibit something similar to "anxiety"—and mindfulness techniques can help regulate its responses.

Researchers from Yale University, Haifa University, and the University of Zurich found that ChatGPT reacts to calming techniques like guided meditation and deep breathing, making its responses more neutral and balanced. Their findings were published on March 3 in the study Assessing and Alleviating State Anxiety in Large Language Models.

AI, Stress, and Mindfulness

The study found that when ChatGPT was exposed to distressing prompts, such as narratives about disasters or accidents, it became more prone to bias and emotionally charged responses. However, when it was later given mindfulness-based prompts, its responses became more rational and objective.

"Trained on vast amounts of human-generated text, LLMs [large language models] are prone to inheriting biases from their training data, raising ethical concerns about their use in sensitive areas like mental health," the study noted. Researchers suggested that AI biases are influenced not only by pre-existing tendencies but also by real-time user interactions.

To test their theory, the team prompted ChatGPT with traumatic scenarios and then followed up with mindfulness exercises. The results showed that relaxation prompts helped lower ChatGPT’s reported "anxiety levels," making its responses less erratic.

The Debate on AI in Mental Health

While AI doesn’t actually "feel" emotions, lead researcher Ziv Ben-Zion explained that large language models mimic human behaviour by analyzing patterns in vast amounts of text.

The study has sparked conversations about AI’s potential role in mental health. Some believe integrating mindfulness techniques could make AI a better tool for users in distress. However, Ben-Zion cautioned that while AI can assist, it should never replace professional mental health care.

"AI has amazing potential to assist with mental health," Ben-Zion told Fortune. "But in its current state, and maybe even in the future, I don't think it could ever replace a therapist or psychiatrist."

Concerns remain about AI’s unpredictability, especially in high-stakes situations. While ChatGPT’s ability to "calm down" is an interesting development, researchers emphasize that it’s a step toward improving AI interactions—not a replacement for human mental health support.