About 80% Follow AI Advice for Personal Problems, but No Increase in Happiness

The research team conducted an experiment with 6,474 adults in the UK. Participants engaged in a 20-minute conversation with AI models such as GPT-4o, Llama-3.3-70B, and Gemini 3 Pro. They were divided into a group that consulted the AI about personal problems and a control group that engaged in casual conversation about hobbies.

As a result of the experiment, 75% to 79% of participants who consulted about personal problems reported that they acted on the AI's advice in the following weeks. This figure exceeds the 55.4% observed in the group that chatted about hobbies.

More than 65% of people followed the AI's advice even when the suggestions involved high risk. The researchers stated that people tend to accept AI's words relatively easily and translate them into real-world action, regardless of the significance of the outcome.

On the other hand, the AI's advice did not provide lasting benefits. In a follow-up survey conducted two to three weeks later, the group that consulted about problems showed no meaningful improvement in happiness compared to the group that engaged in casual conversation.

Immediately after the conversation, the group consulting about problems exhibited lower levels of happiness. However, this difference disappeared after several weeks, and both groups returned to their original levels, according to the report.