Rival AI companies Anthropic and OpenAI have released dueling studies that paint a picture of how people are using their flagship products, ChatGPT and Claude. Both pieces of research analyzed large datasets of user conversations, examining work and non-work-related conversations.

While the two companies used different datasets and methods, OpenAI’s analysis suggests the consumer versions of ChatGPT are used mainly for personal and exploratory purposes, whereas Anthropic’s findings show Claude.ai and Claude API are primarily used for work-related tasks like coding, research, and education.

According to the study released by OpenAI, most ChatGPT conversations aren’t about work at all. Non-work-related messages made up more than 70% of all usage, up from 53% in June 2024, while work-related queries made up 27% of all messages, down from 47% of total conversations a year ago.

The research suggests that ChatGPT is becoming more of a general consumer product than an enterprise tool. The three most common ChatGPT conversation topics were categorized by researchers as practical guidance, writing, and seeking information: these three categories collectively account for nearly 78% of all messages.

However, it is worth noting that OpenAI’s dataset covered usage on consumer ChatGPT Plans (Free, Plus, Pro) and did not include non-consumer plans such as Teams, Enterprise, or Education.

When ChatGPT was used for work, the study found that users appear to derive the most value when using the chatbot like an advisor or research assistant, rather than asking it to perform tasks directly. The researchers argue in the study that ChatGPT boosts worker productivity primarily through decision support. It also found that users in highly-paid professional and technical occupations are more likely to use ChatGPT for work.

The study found that writing tasks, which included editing and drafting, were the most common work use, accounting for 42% of work-related messages and more than half of all messages for users in management and business occupations. Around two-thirds of these requests were to modify existing text rather than create original text from scratch.

The number of people using ChatGPT for coding tasks is even smaller, with only 4.2% of total messages related to computer programming, compared to Claude’s 36%. Technical Help, the umbrella category that included computer programming, also had the lowest apparent user satisfaction of seven categories that the study examined.

In contrast, research from Anthropic found that Claude is used heavily for work-related productivity, especially coding, education, and research.

Mathematical tasks and coding were the dominant activities globally for Claude.ai. Among other domains, the fastest-growing areas are education, which has increased by roughly 35% since December 2024 and now accounts for just shy of 13% of all use, and scientific research, which now represents around 7% of usage. In contrast, traditional office and business tasks have declined: management-related tasks have fallen from 5% to 3%, and business and financial operations have decreased from 6% to 3%.

Businesses, particularly those using Claude through the API, primarily use the tool for automation-heavy work, often for “full task delegation,” with 77% of API tasks automated compared to roughly 50% on Claude.ai. The research suggests businesses are using the technology to automate rather than collaborate on work. These business-focused interactions are concentrated in coding, which accounts for 44% of API use, as well as administrative support. A further 5% of API usage is dedicated to developing or evaluating AI systems.

The dueling studies suggest that users are favoring specific models or products for different types of tasks.

ChatGPT is emerging increasingly as a personal or exploratory tool, used for writing, information-seeking, general advice, and casual interaction, while Claude is a more work-focused productivity tool, used heavily for coding, research, and business automation. For example, Claude has been popular among software engineers for some time. This market split also suggests that different AI companies could be carving out complementary niches rather than directly competing on all fronts.