How People Use ChatGPT (2025): Key Insights
General Trends and Types of ChatGPT Usage
How People Use ChatGPT (2025): Key Insights
- Scale: ~700M WAU • ~18B msgs/wk
- Non-work: ~70% of all messages
- Education: 10.2% tutoring/teaching
- Intent: 49/40/11% Asking/Doing/Expressing
- Top 3: Practical Guidance • Seeking Information • Writing
| Metric | Value | Note |
|---|---|---|
| Weekly Active Users | ~700M | global |
| Messages per Week | ~18B | aggregate |
| Non-work share | ~70% | mid-2025 |
| Education (tutoring) | 10.2% | of all messages |
| Intent split | 49/40/11% | Asking/Doing/Expressing |
Frequently Asked Questions
How large is ChatGPT’s audience in 2025?
What are the top use cases?
What share is educational (tutoring/teaching)?
What is the intent split (Asking/Doing/Expressing)?
Which dominates—personal or work usage?
Who uses it most—young or older users?
Rapid growth and global spread: Since its public debut in November 2022, ChatGPT has achieved extremely rapid and widespread usage. By July 2025, it already had about 700 million active users per week, who were sending a total of 18 billion messages per week – equivalent to ~10% of the world’s adult population[1]. This pace of global adoption is unprecedented for a new technology[1]. Initially, the user base was dominated by men (nearly 80% of early users), but by mid-2025 the gender split had equalized – about 52% of active users were now names typically associated with women [2][2]. The growth in usage has been highly global: over the past year there was especially high growth of users in low- and middle-income countries, suggesting that the service is being widely adopted beyond the most technologically advanced markets[3][3].
Work vs. personal use: The data indicate that although ChatGPT is used for both professional and personal purposes, non-work uses are growing faster. Between June 2024 and June 2025, the average daily number of user messages jumped from ~451 million to ~2.63 billion messages, and the share of personal (non-work) uses rose from about 53% to over 70%[4][4][5]. In other words, by mid-2025 most interactions with ChatGPT were related to education, hobbies, or everyday life, rather than directly to professional duties. (It is important to note that even outside of work the chatbot contributes to productivity – for example by helping in “household production” or personal projects – on a scale comparable to or even greater than its impact in paid work [5].) The decline in the relative share of work-related cases is mainly due to a change in behavior of the users themselves (who over time use it more for personal purposes), and not just an influx of new users [6]. Nevertheless, work usage remains significant and is found to be concentrated among certain groups – for instance, people with higher education and in higher-paying professions use the chatbot for work significantly more often than others [7].
Conversation topics (main categories): The vast majority of user sessions with ChatGPT fall into a few broad topical groups. Nearly 80% of all conversations are classified into three main categories: Practical Guidance, Seeking Information, and Writing [8][5].
- Practical Guidance: This is the most common category, including requests for advice, instruction and ideas – for example help with learning and explanations (acting as a “private tutor”), “how to” instructions for a variety of tasks, advice for personal goals (workouts, health, travel, etc.), or creative idea generation [9].
- Seeking Information: Queries for facts and references, similar to an internet search – e.g. information about people, current events, products, recipes, and so on. These requests aim for exact factual answers, and in this case ChatGPT acts as a more flexible alternative to a web search engine [10], providing a summarized or explained answer to the query.
- Writing: Requests in which the user asks the AI model to generate or edit text. This includes automatically composing emails, documents, articles, as well as editing, correcting, summarizing or translating text provided by the user [11]. This category stands out especially in a professional context – “Writing” is the most common type of task in work conversations, forming ~40% of all work messages in June 2025 [12][6]. Indeed, the ability of models like ChatGPT to generate whole pieces of text (reports, code, etc.) is what distinguishes them from traditional tools.
The above three categories dominate ChatGPT usage, together accounting for around 78% of all user conversations [13]. The remaining share is distributed among more specialized or rarer uses: Technical Help – e.g. requests for assistance with programming, mathematical calculations or data analysis – and Self-Expression – conversations on emotional, personal topics or role-play – along with a few other smaller categories. Interestingly, computer programming constitutes a relatively small share of queries – only ~4.2% of all messages [14][7], despite the popularity of AI among developers. Likewise, “virtual friend” or emotional support chats are a very small niche – ~1.9% of all messages are on the topic of relationships or personal reflections, and ~0.4% are games and role-play [15]. These numbers refute some early claims that chatbots would be used primarily for therapeutic or social purposes – on the contrary, with ChatGPT the pragmatic tasks like information, learning, and assistance at work predominate [16].
Types of Interactions (Asking, Doing, Expressing)
Apart from the conversation topic, the study also classified the user’s intent in each query. Three distinct interaction categories were introduced – Asking, Doing, and Expressing – which describe what the user is trying to achieve with their message [17][18]:
- Asking – when the user is seeking information, an explanation, or advice to become better informed and to make a decision. This corresponds to a consultative role (for example asking for facts, definitions, recommendations) and is equivalent to asking for guidance [19][20].
- Doing – when the user wants a specific task to be executed or some artifact to be produced. This could be generating text (e.g. a letter, code, summary) or another activity where the model primarily produces content on behalf of the user. This category reflects requests that are task execution [21][22].
- Expressing – when the user is simply expressing thoughts, feelings or a viewpoint, without expecting a concrete answer, solution or action. These cases have no clear goal of seeking information or a result – rather, the chatbot serves as a listener or “diary” (e.g. sharing an experience). These are rare cases of “self-expression” without seeking output data or tasks [18].
The distribution among these interaction types shows that users most often use ChatGPT as a source of knowledge and advice. About 49% of all messages fall under Asking (seeking information/advice), nearly 40% – under Doing (assigning a task to perform), and about 11% – under Expressing (sharing without a specific goal) [23][8]. In other words, almost half the time ChatGPT is used for informational support in problem-solving or decision-making, and a significant portion – for directly performing tasks instead of the user. The data also show an interesting dynamic: the share of Asking messages has been growing faster over the past year, outpacing the growth in Doing messages [24]. Users are increasingly posing questions for advice or clarification, and not only tasks for execution – a trend supported by the ratings: Asking-type requests receive higher scores for usefulness and satisfaction compared to Doing [24][25].
It is important to note that the Asking/Doing ratio varies by context. In work situations, users more often use ChatGPT for concrete tasks: ~56% of work-related messages are classified as Doing (primarily various forms of writing), while ~35% are Asking and ~9% are Expressing [26]. This means that in the workplace the chatbot often serves as an instrument for execution – for example generating or formatting textual materials – whereas in a personal context the share of questions (Asking) is higher. In both cases, however, ChatGPT acts predominantly as an assistant and “co-pilot” that supports the user’s thought process, rather than simply as an automaton for ready solutions. This role is evident from the fact that even in work settings most users derive value from ChatGPT as an advisor or research assistant, not only as a tool for mechanically generating final products [27].
ChatGPT Usage by Young People (up to age 26)
Share of young users: The study shows that young people dominate the user base of ChatGPT. Almost 50% of all messages (specifically ~46%) come from users aged 18–25 [28]. In other words, nearly half of active users are youth under 26, which highlights the enormous popularity of the chatbot among the generation of students, university learners, and young professionals. As 2025 progresses, the age differences smooth out slightly (i.e. older people also begin to use the service more), but the youth group continues to generate a significant portion of the traffic [29].
Educational purposes and learning: The high share of young users naturally leads to many instances of using ChatGPT for school, university or self-study. The data confirm that education is a key sphere of use – approximately 1 out of every 10 messages (10.2%) to ChatGPT is a request for an explanation, a lesson or help with study material [30]. In fact, educational queries represent fully 36% of all “Practical Guidance” conversations, which means a significant portion of users are seeking in the chatbot the role of a teacher or instructor [30]. For example, young people often use ChatGPT to have it explain a complex concept, help with solving a problem in math or code, translate a text into plain language, or suggest ideas for projects and essays. This online tutor function makes ChatGPT a valuable tool for learning outside the classroom.
The statistics also reveal one substantial difference: young people use ChatGPT less for professional purposes compared to older users. Among users under 26, only about 23% of their messages are related to work, and this share increases with age [31]. In other words, young people primarily engage ChatGPT for educational and personal needs, while older users (especially above ~30) more often integrate it into their work process. This is expected, since a large portion of 18–25-year-olds are still in education or at the start of their careers. The chatbot has become a helper for doing homework, preparation for exams, searching for academic information and even writing motivation letters or CVs.
Topics and preferred interaction style among youth: Although the data do not segment every topic in detail by age, we can conclude that young users have similar core interests on the chatbot: they ask questions about study content, obtain practical advice (e.g. how to study effectively, how to solve a particular problem), and often take advantage of the model’s ability to generate text – whether it be translating a difficult text, checking essays, creating sample solutions, writing code for school projects, or even generating ideas for creative assignments. One could say that for the generation under 26, ChatGPT serves as a universal learning assistant – combining the functions of an encyclopedia, a mentor and an editor. This trend is especially visible among students and high schoolers who seek quick and understandable explanations on a variety of topics. From the perspective of the Asking/Doing classification, young people likely have a significant share of “Asking” interactions (because they often ask questions to understand), but they also do not hesitate to use the “Doing” functionality – for example, to ask ChatGPT to write a sample passage for them, to translate a segment or to format text.
In summary, young people are a driving force in ChatGPT usage, actively using it in their education and personal development. ChatGPT fills the niche of an accessible 24/7 learning assistant – providing knowledge, explanations and even motivation for learning. Since only about one quarter of their activity is work-related, it can be argued that for this age group the chatbot is above all a tool for learning and self-improvement. (Still, as they enter the workforce, these young users will likely carry their AI usage skills into the professional realm, as hinted by the data on the higher share of work uses among more educated and older groups [7].)
Main Insights from the Research and Comparison with Other Platforms
Unprecedented adoption speed: ChatGPT’s spread has occurred with unprecedented speed – in ~2.5 years it reached ~10% of the world’s adult population [1][1]. This pace surpasses all previous technologies in terms of global adoption [1]. For comparison, by the end of 2024, 28% of Americans had tried ChatGPT [32], more than any other chatbot. In April 2025, ChatGPT had over 10 times more daily users than competing models like Claude or Copilot [33], cementing its leading position in the market.
From work to personal life – changing usage pattern: While early speculation focused heavily on AI’s influence on jobs and productivity, the data shows that non-work uses are growing even faster. The share of personal conversations (learning, hobbies, everyday life) increased from ~53% in 2024 to over 70% in 2025 [5][4]. This means that ChatGPT is entering people’s everyday lives on a massive scale, and is not confined only to the office. The benefit outside of work – for example assistance at home, in personal projects, for self-improvement – may be equal to or even greater than its impact on paid employment [5]. This finding complements economic analyses focused predominantly on the labor market, by showing that generative AI brings significant consumer welfare. (According to estimates by Collis and Brynjolfsson, U.S. consumers would forgo generative AI only in exchange for compensation of ~$98 per month, which implies an annual consumer surplus of over $97 billion [34][9].)
Three dominant usage scenarios: Globally, the main use cases are grouped into practical advice, information seeking and writing – nearly 80% of all chats in total [13]. This reflects ChatGPT’s role as a universal assistant – from answering questions and giving guidance, to generating and editing text. Writing stands out especially, since almost every profession requires good written communication. As a result, “Writing” tasks prevail in workplace settings (around 40–42% of work messages) [12][6], which underscores the chatbot’s unique ability to create and polish a variety of digital content.
ChatGPT as a tool for decisions and advice, not just automation: One of the key ideas in the report is that users value ChatGPT mostly as support for decision-making and knowledge expansion, rather than simply as a machine for ready answers. Almost half of all queries (49%) are of type Asking – seeking clarifications, recommendations or information [23]. This accords with the finding that “providing information and help with decisions are the most frequent use cases” of the model [35]. Users use it as a consultant or “co-pilot,” which improves their own thinking and productivity, especially in knowledge-intensive activities [36]. This model – AI as advisor – differs from the notion that AI will entirely replace human labor; instead, people integrate it as an intellectual partner in their tasks.
High value in knowledge-intensive professions: Analysis by occupation shows that in almost all fields ChatGPT is used to improve access to information and the quality of decisions. Around 58% of work messages relate to activities like “gathering and interpreting information” or “making decisions, giving advice and solving problems” – tasks common to numerous professions [37][35]. This broad usage pattern – whether in management, engineering, administration or sales – shows that ChatGPT finds a place in all levels of “white-collar” work, mainly as a tool for problem-solving and idea generation. This corresponds to the theory that AI boosts productivity most by improving the quality of human decisions (and not only by automating routine tasks) [36].
Significant share of “editorial” tasks: Approximately two-thirds of all writing tasks on ChatGPT are editing or further developing text provided by the user, rather than completely new writing [12]. Users often give their own text (an email, essay, description, etc.) and ask the model to improve it – to make it more formal, clearer, to summarize the key points, to translate it or check the grammar. This underscores the role of AI as an “intelligent pen” that fine-tunes and polishes human written expression, instead of fully replacing it.
Modest shares for programming and “chat for companionship”: Contrary to some enthusiasts’ expectations, coding occupies a relatively modest share of ChatGPT usage – only ~4.2% of all messages are related to programming [14][7]. This is much lower than the data for the competing model Claude, where ~33% of work conversations are about code [14]. The difference likely stems from the different user profiles – ChatGPT has a much broader mass audience, whereas Claude (analyzed by Handa et al., 2025) is used more in professional contexts by specialists. Also, although the idea of a chatbot as a “virtual friend” or therapist is often mentioned, only around 2% of conversations with ChatGPT are on topics like relationships, personal feelings or role-play [15]. This contrasts with one report (Zao-Sanders, 2025), which by manually collecting online opinions concluded that “therapy/companionship” is a leading use case of gen. AI – evidently, the real data do not support this conclusion [38][10]. ChatGPT is used primarily for useful tasks, while its emotional-social applications remain niche.
Closing “digital divides”: Over time, some initial differences in the profile of ChatGPT users have been blurring. Gender is no longer a factor – if in the first months ~80% of users were men, by mid-2025 active women slightly outnumber men (52% vs 48%) [2]. The age distribution also shows that older groups are catching up in usage, though young people (18–25) remain the most active [29]. The geographic aspect is also interesting: developing economies registered the highest growth – many middle- and lower-income countries saw a jump in ChatGPT adoption from 2024 to 2025, reaching similar or higher usage levels compared to wealthy nations [39][40]. This suggests that generative AI tools are quickly being democratized on a global scale, offering access to knowledge and assistance everywhere that has internet.
More educated users integrate AI more into work: Not surprisingly, people with higher education and professionals in well-paid fields are more inclined to integrate ChatGPT into their work process [7]. Among users with an educational level above the average, the share of messages that are for work purposes is noticeably greater (for example, ~46–48% for college-educated users, versus ~37% for those with lower education) [41][11]. Moreover, educated users more often use the “Asking” mode at work, i.e. they pose questions for advice, research and solutions, instead of directly delegating tasks for execution [42]. This supports the thesis that AI increases returns for experts, acting as an “intelligent collaborator” that helps them make better decisions and be more productive.
In summary, the main findings of the study paint a picture of ChatGPT as a widely used tool, diverse in its applications and user base, which has a noticeable impact in both professional and personal realms. Contrary to some expectations, actual usage is concentrated in pragmatic tasks – seeking knowledge, advice and assistance in content creation – rather than in technical or emotional niche cases. Compared to other platforms, ChatGPT has a broader and more general usage profile (less specialized in programming than Claude, for example) and a far larger scale of audience. These insights are valuable not only for understanding the current influence of AI, but also for predicting the future social and economic effects of the spread of such technologies.
Most Common Categories and Topics (Key Phrase Analysis)
As became clear, most user sessions with ChatGPT can be grouped by thematic orientation. The largest share consists of requests for practical advice and information seeking – these are essentially decision support situations where the user wants to learn something new or get guidance in a given situation. Next are tasks related to writing and text editing, which are also extremely frequent. Below the main categories of tasks and topics are summarized, with examples and their share of all messages:
| Category / Task | Examples of queries | Share of all messages |
| Educational queries (Tutoring/Teaching) | explanation of study material, help with homework | 10.2% [30] |
| General “How to…” advice | practical guidance in various areas (e.g. “how to learn programming”, “how to plan a budget”) | 8.5% [43] |
| Programming (Technical Help – Coding) | writing code, debugging, algorithmic tasks | 4.2% [44] |
| Mathematical calculations | solving math problems, formulas, equations | 3.0% [44] |
| Data analysis | statistical analysis, interpreting data sets | 0.4% [44] |
| Relationships and personal reflections | relationship advice, sharing personal issues | 1.9% [45] |
| Games and role-play | fictional scenarios, chat “for fun” | 0.4% [46] |
Table: Key thematic subcategories and their approximate share of the total number of user messages.
The above categories represent both the most popular and some more rarely encountered uses. Education and “how to” advice clearly stand out – together over 18% of all interactions – which shows how valuable ChatGPT is as a teacher and consultant in everyday life. Likewise, technical assistance (especially programming) has its not-insignificant share, though not the leading one, and personal-emotional topics occupy a very small part of the overall volume, as discussed.
Another important spectrum of tasks are the various types of written activities, grouped in the category Writing. Within this category, a few typical phrasings of user requests are noticeable (arranged from most to less common): editing or critique of provided text, personal correspondence or documents (e.g. emails, letters), translation of text, generation of arguments or summaries, and writing creative text [47][48]. Three of these five types – namely editing text, translation, and summarization/argumentation – are cases in which the user gives the model some initial text and wants it improved or reworked. These sub-tasks make up about 2/3 of all “writing” conversations [49]. The remaining two subtypes (creating entirely new text – personal communication from scratch or creative writing) form about 1/3. This ratio again confirms that ChatGPT most often acts as an editor and assistant – the user already has content or an idea, and the model helps to polish or develop it. For example, among frequently seen phrases in requests are “Rewrite the following text to sound more formal,” “Summarize these paragraphs for me,” “Translate into English and fix the style,” “Finish my story,” etc. – all of which imply that the model needs to take some input and transform it in a useful way.
Regarding information seeking, popular keywords include everything related to “who/what/when/how” – users ask for facts (e.g., “Who was President after Lincoln?”, “What is the capital of …?”, “Why is the sky blue?”, etc.), explanations of scientific or societal phenomena, current events, product or destination recommendations, recipes, and more. These queries are often formulated similarly to a standard Google search, but users expect a summarized and comprehensible answer, without having to themselves sift through information from multiple sources. ChatGPT satisfies this need by providing a direct answer or synthesis on the given topic.
Practical advice usually begins with phrases like “How to…” or “Give me ideas for…”. For example: “How can I improve my presentation skills?”, “How to start my own business?”, “Ideas for a healthy dinner for 4 people”, etc. – this type of query positions ChatGPT as an expert advisor in various fields, from health and fitness to career, finance, and hobbies. The model responds with concrete steps, recommendations or creative ideas, often personalized to the description the user has given (for example tailored to their experience or goals) [50]. It is precisely this personalization and ability to generate new content (rather than only finding something existing) that differentiate ChatGPT from ordinary search engines [51][52].
Technical help, though more limited in share, also has characteristic key phrases: “Write a function that…”, “How do I fix this error in the code?”, “Explain this algorithm to me”, “Solve this statistics problem.” Users often provide parts of code or data and ask ChatGPT for debugging, optimization or interpretation of results. In math queries – they give an equation or problem and want a step-by-step solution. Here ChatGPT acts as a combination of mentor and calculator, providing not only the answer but also an explanation of the steps.
Finally, though rare, there are also requests for fun or creativity – for example generating short stories, poems, jokes, or engaging in a role-play (“Imagine you are a character from a game and…”). These cases (covered in the Self-Expression or Games/Role Play category) constitute a small percentage (around 2–3% in total) [53], but demonstrate the model’s flexibility to engage in creative, non-utilitarian conversations.
In summary, the lexical and thematic analysis of user queries to ChatGPT highlights several key motives: users seek knowledge, advice, help in creating or improving text, solving specific problems (technical, educational, logical), and sometimes just creative interaction. The most frequently used words and phrases reflect this – questions like “how”, “why”, “write”, “explain”, “help me to” are ubiquitous. ChatGPT, judging by the data, has established itself as a versatile tool that people use for learning, writing, planning and decision-making in their daily lives. This makes it not just a chatbot for entertainment, but rather a universal digital assistant, capable of tuning itself to the needs – be it of a student, a professional or a curious mind in free time.
Sources: The findings and data are based on the report “How People Use ChatGPT” (OpenAI, September 2025) [8][12][14] and related analyses. All percentages, statistics and statements in the text are supported by the results in this document.
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] economic-research-chatgpt-usage-paper.pdf
