OpenAI Study Uncovers 3 Surprises About How People Use ChatGPT

ChatGPT is playing a growing role in decision-making at work and at home, according to a new study released this week by OpenAI. Nearly three years after its public debut, the chatbot has become a go-to tool for 700 million people each week, driven by a surge in women users and a sharp rise in personal queries that far outweigh work-related ones.

The findings come from a working paper published by the National Bureau of Economic Research (NBER), authored by OpenAI’s Economic Research team and Harvard economist David Deming. The study analyzed 1.5 million ChatGPT conversations using a privacy-preserving methodology, where “no user messages were observed by humans during any part of the work on this paper,” OpenAI said.

The research highlights how user behavior is shifting as generative A.I. adoption widens. Here are three surprising findings:

Personal use outpaces work use.

Despite concerns that A.I. adoption could replace jobs, most ChatGPT queries are personal in nature (for now). The study found that 70 percent of inputs relate to everyday life, compared with just 30 percent for professional use.

Examples include people asking for explanations, practical guidance or self-improvement tips. While businesses continue to integrate generative A.I. into the workplace, the paper suggests the most immediate impact is how individuals turn to ChatGPT in their daily routines.

Women now make up a majority of users.

Generative A.I. platforms were once dominated by male early adopters, but the balance has shifted. As of July 2025, 52 percent of ChatGPT’s weekly active users had first names the researchers classified as typically female, suggesting usage now reflects the general adult population. “Early adopters were disproportionately male, but the gender gap has narrowed dramatically,” the report notes.

That marks a break from findings in a 2024 study by Berkeley Haas, Stanford and Harvard Business School researchers, which reported a persistent male skew across leading chatbots, including ChatGPT, Perplexity and Claude. Those researchers warned that underrepresentation of women could worsen bias in A.I. systems by skewing training data. If the gender gap continues to close, the new study suggests, those risks may be reduced.

Decision-making support is rising as writing help declines.

ChatGPT is increasingly being used as a guide rather than a ghostwriter. In July 2024, writing assistance accounted for 36 percent of queries; by July 2025, that figure had dropped to 24 percent. Meanwhile, “practical guidance” queries, such as how-to advice, tutoring, creative ideation and self-care, grew to nearly 29 percent of usage.

In the workplace, ChatGPT is being deployed as a research assistant or advisor, helping knowledge workers make complex decisions. That trend has raised concerns among experts who caution against overreliance. “The overreliance on A.I. can result in complacent behavior and reduce critical thinking, which is essential for decision-making,” wrote Kamales Lardi, CEO of Lardi & Partner Consulting and author of Artificial Intelligence for Business (2025). She added, “Exposure to diverse viewpoints or information results in skewed decisions, restraining the human brain’s natural ability to adapt and question biases during the decision-making process.”