How people are using ChatGPT

Source: OpenAI Date: 2025-09-15 URL: https://openai.com/index/how-people-are-using-chatgpt

Summary

OpenAI research/blog post from September 2025 analyzing usage patterns across the ChatGPT user base — likely based on aggregate, anonymized query analysis. Published in the period immediately following the PBC restructuring, this is partly an accountability document (here’s what the product is actually being used for) and partly marketing (look at the beneficial uses). The timing with the nonprofit/PBC statement suggests OpenAI wanted to demonstrate social utility alongside governance changes.

Implications

Usage data as mission evidence. OpenAI’s public benefit mission requires demonstrating that ChatGPT benefits users broadly, not just paying subscribers. An aggregate usage analysis showing healthcare questions, education assistance, and professional productivity — rather than primarily entertainment or NSFW content — supports the mission narrative. The selection of what to highlight in such a post is itself a communications decision.

What usage data reveals and conceals. Query-level analysis can show topic distribution but rarely reveals whether the answers were accurate, whether they caused harm, or whether users acted on AI outputs in ways that affected them. A “how people use ChatGPT” study that only covers query topics without outcome data is partial evidence of impact.

Thread: usage and impact research. Sits alongside the HealthBench evaluation (May 2025) and the state of enterprise AI report (December 2025) as OpenAI’s published evidence base for its social impact claims. The usage research is consumer-facing; the enterprise report is B2B-facing.

Watch: Whether the usage patterns described in September 2025 shift measurably after the advertising introduction (January 2026) and the ChatGPT Go launch — advertising and broader free access change the user composition in ways that could show in usage pattern analysis.