With around 700 million weekly active users and 18 billion messages sent per week, ChatGPT has become a piece of global infrastructure. Yet beyond the raw user numbers, it long remained unclear what exactly these people are actually doing. A new Working Paper (“How People Use ChatGPT”), analyzing data from May 2024 to July 2025, now provides the first empirically reliable answer.
The results are surprising and correct common narratives: ChatGPT is neither primarily a programming assistant for the tech elite nor a virtual friend for the lonely. Instead, the image emerges of a universal tool for information processing. The analysis identifies four central patterns that dominate the global use of this AI.
1. The “Big Three”: 80 Percent of Usage
The study’s arguably most important finding is the enormous concentration of usage. Around 80 percent of all conversations worldwide fall into just three thematic categories. This clarifies that for the broad masses, ChatGPT is primarily a pragmatic “tool for the mind”:
- Practical Guidance: This is the largest block. Users seek individualized advice, have complex issues explained to them, or use the AI for brainstorming and idea generation.
- Seeking Information: Here, the chatbot replaces or supplements the classic search engine. It is about quickly finding specific facts in a dialogue format.
- Writing: Support in the creation, structuring, and refinement of texts.
2. Writing Reigns Supreme in the Work Context
Looking at work-related usage in isolation, writing emerges as the “killer app” of generative AI. About 40 percent of all professionally motivated messages in June 2025 revolved around text work. The data reveal an interesting detail here that contradicts the cliché of the “AI that writes everything on its own.”
Two-thirds of these writing tasks do not consist of the AI generating a text from scratch. Instead, professionals use ChatGPT to revise, summarize, translate, or stylistically polish texts provided by humans. Thus, in everyday work life, the AI functions less as an author and more as an editor and copy editor, increasing efficiency in daily text production.
3. Information Processing as a Core Activity
The study mapped the messages to the standardized work activities of the O*NET system. The result is clear: ChatGPT is an instrument for managing information overload. The three most frequent activities, which together account for almost half of usage, are:
- Getting Information: Obtaining relevant data.
- Interpreting Meaning: Understanding and categorizing information for others.
- Documenting/Recording: Capturing and structuring knowledge.
Especially in knowledge-intensive professions, the AI serves as a cognitive co-pilot. It supports decision-making processes (“Making Decisions and Solving Problems”) by processing information rather than automating physical routine tasks. The intent “Asking” (questioning/understanding) increases over time and shows the highest user satisfaction, underscoring its role as a knowledge assistant.
4. The Myth of Coding and Social Replacement
Just as revealing as what people do is what they hardly do. The data set straight two widespread media narratives:
First, ChatGPT is not purely a tool for software developers. Programming tasks (“Coding”) make up only about 4 percent of the total message volume. While this figure is higher in technical professions, it remains a niche in the overall view.
Second, dystopias in which AI replaces human relationships are not confirmed by the data. Usage for purely social conversations, role-playing, or as a partner substitute lies significantly below 3 percent. People use ChatGPT predominantly functionally and purposefully—whether for work or private projects—but hardly as an emotional surrogate partner.
Conclusion
The empirical reality of ChatGPT is less science fiction and more a massive productivity boost in daily knowledge work. The world uses the system primarily to tame information and process texts. It has established itself as a universal interface for knowledge that complements human capabilities in writing and decision-making, rather than replacing them through autonomous actions or simulated feelings.
