700 million people send a staggering 18 billion messages per week. OpenAI reveals the most popular use of ChatGPT for the first time.

[Introduction] The first usage report of ChatGPT has been released! Its weekly and monthly active users have soared to 700 million. It has become an essential office tool for white - collar workers with high education, while programming has become less popular. Meanwhile, the latest report from Anthropic shows that the proportion of tasks people assign to Claude has skyrocketed to 49%.

The first usage report of ChatGPT is out!

Today, OpenAI, in collaboration with Duke and Harvard, released "How People Use ChatGPT", a report with a total of 63 pages.

Report link: https://cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f142e/economic-research-chatgpt-usage-paper.pdf

From its launch in November 2022 to July 2025, the report covers a vast amount of data over nearly three years, focusing on analyzing the usage of ChatGPT.

As of July, the weekly active users of ChatGPT have exceeded 700 million, accounting for about 10% of the world's total adult population, with a total of 18 billion messages per week.

· Who is using ChatGPT?

· What do people use ChatGPT for?

Generally speaking, the growth rate of non - work - related information is faster, soaring from 53% to 70%.

Among them, practical advice, information query, and document writing are the three most common themes, accounting for 80% of the conversations.

The gender gap in the user profile is also narrowing, with an increasing proportion of female users.

Notably, the report also reveals that people rarely use ChatGPT for programming and self - expression.

At the same time, Anthropic released a brand - new AI economic index report, along with an interactive website for exploration.

Interactive website link: https://www.anthropic.com/economic-index#job-explorer

In the report, the most significant finding is that the proportion of people assigning complete tasks to Claude has increased from 27% eight months ago to 39%.

Moreover, Washington DC, Utah, and California have the top three per - capita Claude usage rates in the United States.

Different from ChatGPT, Claude's main battlefield is code writing, which is also one of the most frequently used scenarios (accounting for about 36%).

The report states that the proportion of tasks involving new code has more than doubled. Developers spend less time debugging and modifying code, which indirectly confirms that Claude is becoming more and more reliable.

Let's break down the details of the "report cards" of these two major AI tools one by one.

ChatGPT's First Report: Programmers Flee in Droves

OpenAI's research adopted a more scientific method, which is based on internal conversation data and uses automated classification combined with privacy - protection technology for analysis.

1.5 million conversations cover three major datasets. Specifically, they include:

Growth data: From November 2022 to September 2025, the total daily messages of all paying users and basic demographic information

Classified message data: From May 2024 to June 2025, about 1 million randomly selected de - identified messages, coarsely classified

Employment data: Aggregated from the publicly available occupation and education information of about 130,000 users

ChatGPT Timeline

It can be seen that since ChatGPT's debut, the number of weekly active users has been soaring, with an increase of more than 30 billion this year alone. Meanwhile, the total daily messages have also been continuously rising.

In the following figure, it can also be seen that early - registered users maintain a higher usage rate. However, the overall trend is on the rise.

On the one hand, it is related to the continuous improvement of the model's capabilities; on the other hand, people have discovered new application scenarios for existing uses.

As mentioned before, the report mainly focuses on two core questions: Who is using ChatGPT? What do they use ChatGPT for?

Real Use: Querying and Asking Questions

Most people use ChatGPT mainly to complete daily tasks.

Among them, 75% of the conversations revolve around practical guidance, information query, and writing. In particular, writing accounts for the largest proportion.

As shown in Table 1 below, from June 2024 to June 2025, the proportion of non - work messages has increased the most, from 53% to 73%.

These conversations specifically involve seven major themes and 24 categories.

In the most common writing category, people usually ask to: modify or comment on a specified text, engage in personal writing or communication, translate, generate an argument or summary, and create novels.

In Figure 7, the demand for "practical guidance" has remained stable at 29%, the proportion of "writing" has dropped to 24% (36% last year), and the topic of "information query" has skyrocketed by 10%, accounting for 24% this year.

Interestingly, the proportion of technical questions, such as programming, has dropped sharply from 12% to 5%. OpenAI analysis shows that many developers directly call APIs and other AI tools to write code, such as Codex.

In work - related conversations, writing is still the most commonly used by office workers, accounting for 40%; practical guidance ranks second, and the proportion of technical assistance has dropped from 18% last year to 10%.

As shown in Figure 9 below, the seven major conversation themes are further divided, and the total number of messages in a year is also counted.

Most people use ChatGPT for writing mainly to modify and optimize existing content, rather than pure creation.

Let's talk about education, one of the important application scenarios of ChatGPT:

10.2% of user messages and 36% of practical guidance are mainly used for tutoring and teaching

8.5% of the total message volume is operation guidance on various topics

In the technical assistance part, computer programming accounts for 4.2%, mathematical calculation accounts for 3%, and data analysis accounts for 0.4%.

The proportion of self - expression topics is even smaller, only 2.4%.

The usage patterns of ChatGPT can be divided into three categories: asking questions (Asking), performing tasks (Doing), and expressing oneself (Expressing).

Among them, about 49% of the messages belong to "asking questions", indicating that people are more willing to use ChatGPT as an advisor.

"Performing tasks" accounts for 40%, including text writing, planning, programming and other task - based interactions, about one - third of which are used in work scenarios; "Expressing oneself" accounts for 11%, mainly involving non - task - based interactions such as personal reflection, exploration, and entertainment.

The following figure shows the proportion of asking questions, performing tasks, and expressing oneself in 1.1 million sampled conversations in work and non - work scenarios.