How People Actually Use ChatGPT — Deep research, short, and actionable

Headline findings

People primarily use ChatGPT to think through things — asking for information, judgement, and advice — rather than only to generate finished outputs. About 49% of messages fall into this “Asking” category. NBER+1
“Doing” (task-oriented output like drafting emails, writing, coding) is still large — ~40% of messages — but growing more slowly than Asking. “Expressing” (journaling, roleplay, emotional/creative play) makes up ~11%. NBER
ChatGPT’s user base is broadening: gender gaps have shrunk and the platform now shows more balanced gender representation (about 52% users with typically feminine names) and strong growth among younger people and users in low- and middle-income countries. OpenAI+1
Workplace use has declined proportionally: work-related messages fell from nearly half of usage to roughly a quarter-to-third in the study period — suggesting ChatGPT is increasingly used for personal life and decision support as much as for productivity at work. MarketWatch

The Asking / Doing / Expressing framework — why it matters

OpenAI groups user messages into three plain-English intents:

Asking — requests for information, advice, judgment, decisions, or explanation. Example: “Should I take job A or job B?”
Doing — requests to produce concrete artifacts: write a resume, draft code, translate a paragraph, create a lesson plan.
Expressing — personal reflections, role-play, creativity, emotional venting, or entertainment.

This taxonomy matters because it shifts how we think about value: ChatGPT is not only a “productivity plugin” (doing work for you) but increasingly a thinking partner — a place people go to clarify, weigh options, or brainstorm. That helps explain why satisfaction scores are higher for Asking-type interactions: people value perspective and reasoning support. OpenAI+1

Key statistics you should remember

(These are the most load-bearing numbers from the study and subsequent reporting.)

Asking: ~49% of messages. NBER
Doing: ~40% of messages. NBER
Expressing: ~11% of messages. NBER
ChatGPT weekly/active reach reported in news coverage: hundreds of millions of weekly users (OpenAI reported numbers like ~700M weekly users in related coverage). The Washington Post+1
Demographic shift: proportion of users with typically feminine names rose to ~52% by mid-2025. OpenAI

What people actually do inside each bucket

Asking: fact checks, quick research, comparing choices (jobs, products, study plans), “what should I do?” style decision support, and condensed explanations of complex topics. People use ChatGPT similarly to how they use a search engine — but with more interactive follow-up (ask again, refine, test assumptions). OpenAI+1
Doing: writing support (editing, drafting emails, blog outlines), planning (travel, study schedules), light programming tasks and debugging, and content creation (social posts, scripts). Writing help is one of the largest single task categories. Ars Technica+1
Expressing: creative writing, role-play, personal reflection, low-stakes conversation, and entertainment. While smaller numerically, Expressing is important for adoption and retention: people use ChatGPT as a companionable tool — not a sterile API. Towards AI

What’s changed over time — the trendlines

From novelty to utility: Early adopters used ChatGPT for coding or tech curiosities. The platform has matured into everyday personal and professional assistance. Usage volumes have exploded — from hundreds of millions to billions of messages daily across the platform. MarketWatch+1
Personal beats professional (relatively): The share of non-work interactions has increased. People increasingly rely on ChatGPT for life decisions, learning, and curiosity, not just workplace automation. MarketWatch
Geographic and demographic diffusion: Faster growth in lower-income countries and broader gender balance suggest the tool is spreading beyond early tech-savvy, male-dominated cohorts. That changes what content and features are most useful. OpenAI+1

Watch our Short

Why Asking is growing faster than Doing

Three plausible drivers from the study and analysis:

Lower effort, higher trust: People can ask an open question quickly and iterate. That style maps well to conversational interfaces. OpenAI
Decision support adds unique value: AI’s ability to summarize trade-offs, highlight unseen angles, and propose frameworks for thinking makes it helpful beyond “write this email.” Users appreciate the advisory role. Ars Technica
Search substitution: For many information queries, ChatGPT offers a conversational alternative to search engines, especially when users want concise, tailored explanations. This competes with traditional search for many queries. MarketWatch

Limitations & risks the study highlights (and what to watch)

Accuracy / hallucination risk: ChatGPT can confidently present incorrect or unsupported claims. When people use it for decision-making, the risk of being misled matters. OpenAI and others point to hallucination as an ongoing issue. Users must verify critical facts. OpenAI+1
Privacy and sensitivity of queries: As people increasingly use AI for personal and emotional matters, systems must protect sensitive information and offer safe, ethical responses. OpenAI
Uneven capabilities across tasks: Coding remains a smaller fraction of use than many expect — advanced developer workflows still lean on specialized tools and human oversight. MarketWatch

Practical takeaways — for users, builders, and policy makers

For users

Use ChatGPT as a thinking partner, not an oracle. Start with Ask-style prompts to explore options, then switch to Doing for concrete drafts. Always cross-check facts for important decisions. OpenAI
Iterate conversationally. Break complicated problems into sequences: ask for frameworks, then ask for concrete steps, then ask for draft outputs. This mirrors how the model is being used successfully by many. NBER

For developers and product teams

Design for the Asking flow. Build interfaces that make follow-up easy (versions, compare views, “why this?” explanations). Thoughtful UI for decision support will unlock more value than single-shot generation. OpenAI
Measure satisfaction by outcome, not time-on-task. Users value getting to a decision or clear understanding quickly. Optimize metrics around resolution and correctness. OpenAI

For policy makers and institutions

Regulate around high-risk use-cases. The study shows people use ChatGPT for consequential decisions — financial, health-adjacent, legal-adjacent. Policies should require provenance, transparency, and clear disclaimers for high-stakes outputs. NBER
Support digital literacy. As AI diffuses globally, empowering users to evaluate AI outputs becomes a public good.

Final thought — AI as a companion for thinking

OpenAI’s study reframes the conversation about conversational AI: it’s not just a replacement for rote work or a faster Google. For many users, ChatGPT has become a partner for thought — a place to test ideas, clarify values, and move from confusion to a usable plan. That’s a different and broader kind of economic and social value: helping people make better decisions, learn faster, and get unstuck.

Connect with Us

We encourage you to explore our platform and connect with us on social media. Join the conversation, share insights, and stay updated:

Be Part of the ZAQ.World Journey

ZAQ.World is more than just a platform — it’s a community. By exploring our domains, sharing feedback, and interacting with our content, you join a growing network of individuals who value knowledge, awareness, and opportunity. We invite you to be a part of our mission to simplify learning, updates, and opportunities in one professional space.

Welcome to ZAQ.World — your trusted hub for education, updates, and future possibilities.

Search This Blog

Welcome To The World Of ZAQ