How People Use ChatGPT — NBER large-scale study (w34255)
A new NBER working paper (No. 34255), produced in collaboration with researchers from OpenAI and Harvard, analyzes a representative sample of about 1.5 million anonymized ChatGPT conversations to understand how people use the model in everyday life and at work.
Quick summary
- Dataset: ~1.5 million anonymized conversations (representative sample).
- Main use cases: practical guidance (tutoring/how‑to/idea generation), information‑seeking (research, facts, recipes), and writing tasks (drafting, editing, summarizing, translating).
- Work vs. non‑work: Over 70% of use was non‑work related by July 2025; writing dominates work tasks (~30% of total usage).
- Interaction types classified as “Asking” (~49%), “Doing” (~40%), and “Expressing” (~11%).
- Economic implications: substantial decision‑support value, especially in knowledge‑intensive tasks; faster uptake in lower‑income countries and narrowing gender gaps among users.
Method & links
The authors analyze anonymized convo logs and usage patterns to draw these conclusions. For the full paper and methodology, see the NBER working paper page and PDF:
- NBER page: https://www.nber.org/papers/w34255
- Direct PDF: https://www.nber.org/system/files/working_papers/w34255/w34255.pdf
- OpenAI summary: https://openai.com/index/how-people-are-using-chatgpt/
Citation
NBER Working Paper No. 34255 (September 2025). See the NBER page for full author list and publication details.
Discussion
How do you use ChatGPT in your daily life or work? Share examples or tips in the comments below.