|
Good morning. Three years into the generative AI revolution, researchers are creating a more vivid and detailed picture of how people use leading models.
New studies from OpenAI and Anthropic provide insights into their users. There are no major surprises here, but the results are fascinating nonetheless, given the nuance of the analysis and their hints about how usage may evolve.
The OpenAI study of consumer messages released today found that consumers use ChatGPT for "practical guidance," writing and "seeking information," which accounted for 78% of all messages studied. Coding and what OpenAI termed as "relationships and personal reflection," accounted for 4.2% and 1.9% respectively.
How they did it. OpenAI said it uniformly sampled approximately 1.1 million conversations, and then sampled one message within each conversation. It included messages from May 2024 to July 2025
Assessing value. OpenAI found that ChatGPT likely improves worker output by providing decision support, especially important “in knowledge-intensive jobs where better decision-making increases productivity.” Asking is relatively more common among educated users employed in highly paid, professional occupations.
The company said its findings were consistent with other research showing that AI agents can “serve either as co-workers that produce output or as co-pilots that give advice and improve the productivity of human problem-solving.”
Both types of queries grew rapidly during the 12-period study, but non-work-related messages grew faster. OpenAI said that in June 2024, 53% of messages were not related to work. That number climbed to 73% by June 2025, the company said.
Business on the horizon. Anthropic says businesses use its Claude AI for coding and administrative tasks, Bloomberg reports, citing a new study from the AI startup. Some 77% of company usage was dedicated to automation. Says Anthropic: “We find that API users are significantly more likely to automate tasks with Claude than consumers are, which suggests that major labor market implications could be on the horizon.”
Anthropic also found that users in higher-income countries were not only more likely to use Clause, but “more likely to seek collaboration rather than automation, and more likely to pursue a breadth of uses beyond coding.” It also reported that businesses were more likely than consumers to trust Clause with agency and autonomy.
How are you putting generative AI models to use? Use the links at the end of this email and let us know. We will put this question and more directly to members of the WSJ Technology Council, who gather today and tomorrow in New York for our fall summit. It will be a great conversation. Watch this space for our in-depth coverage.
|
