In September 2025, OpenAI, Duke University, and Harvard University published a paper called “How People Use ChatGPT.” Even months later, I still think it is worth reading because it points toward a deeper shift in software: users are moving from tools toward outcomes.
That shift matters for SaaS companies, indie builders, and AI-native products. It also connects with EasyGlobe’s work around LLM optimization, because the way people use AI changes how products are discovered and evaluated.
For a public summary of the research context, see this report on how people use ChatGPT.
1. AI usage is already massive and global
The source article highlights the scale reported in the study: by July 2025, ChatGPT had around 700 million weekly active users and more than 18 billion weekly messages. That is not a niche tool behavior; it is a mass internet behavior.
- The early gender gap narrowed significantly over time.
- Younger adults made up a large share of usage.
- Growth was especially strong in middle- and lower-income countries.
2. Non-work use grew faster than work use
Many AI business discussions focus on workplace productivity, but the research shows that non-work use grew faster. This matters because the next major AI products may not look like classic enterprise SaaS. They may solve daily life problems, learning needs, planning, writing, and decision support.
This is important for founders because consumer behavior often arrives before enterprise procurement. If users already treat AI as a tutor, planner, editor, and search layer at home, they will expect business software to feel equally direct at work.
3. What do people actually ask AI to do?
The report groups a large share of conversations into three practical categories: guidance, information seeking, and writing. That is important because it shows that AI is not only a coding assistant. It is becoming a general thinking, searching, and editing layer.
- Practical guidance includes tutoring, advice, creative planning, and everyday decision support.
- Information seeking makes AI feel like an alternative to traditional search.
- Writing is especially common at work, but much of it is editing and transforming existing text rather than creating from nothing.
4. Asking, doing, and expressing reveal product opportunities
The study also separates user intent into asking, doing, and expressing. That framing is useful for product builders. A product that only automates tasks may miss users who want advice, interpretation, reflection, or a better way to articulate what they already know.
5. What does this mean for SaaS?
- Products need to sell outcomes, not only dashboards and workflows.
- AI-native products should make the next decision or next action easier, not just expose more controls.
- Data, context, and distribution become harder to copy than a generic interface.
- The most defensible products may combine workflow, proprietary context, and trusted execution.
What should you read next?
Continue with LLM optimization, AI blog, and SEO optimization. For source checks, use Business Insider research summary and OpenAI research.
How should you apply this guide?
Do not treat this as a passive reading note. Turn the article into a small checklist: confirm search intent, define the source of truth, add internal links, check the canonical URL, review image alt text, and verify the production URL after publishing. That habit makes the article useful as part of an operating workflow rather than a one-time content asset.
FAQ
Does the research mean SaaS is dead?
No. It means SaaS products need to move closer to outcomes and decision support instead of only offering tools.
Is coding the main AI use case?
No. The source article notes that coding is a small share compared with guidance, information seeking, and writing.
What should builders do first?
Start by identifying the user outcome your product owns, then design the AI workflow around that outcome instead of adding AI as a surface feature.