How Are we using ChatGPT?
Consumer ChatGPT Usage Study
ChatGPT adoption covers 10% of the world’s adult population…
The recent usage study for ChatGPT showed that non-work related use has dramatically jumped over the past year…what worries me is that in June 2024 virtually 50% of all use was work related.
By July 2025, 18 billion messages were being sent each week by 700 million users, representing around 10% of the global adult population.
There has been much talk on data and model sovereignty…in this case data sovereignty comes to mind.
And how much of work related data is being sent to ChatGPT?
The study also state that work usage is more common for educated users in highly-paid professional occupations. Talk about valuable work data…
The most common topics for use is practical guidance, information retrieval and writing assistance.
These three topics collectively account for nearly 80% of all conversations.
The study analysed over a million ChatGPT conversations from March 2023, finding that only 4.2% of messages relate to computer programming tasks.
In contrast, it references a report that 33% of work-related Claude conversations map to “computer and mathematical” occupations, with 30% or more of imputed tasks being programming or IT-related.
This higher proportional focus on coding with Claude is attributed to differences in user bases, where Claude appears to attract a greater share of technical or developer-oriented users compared to ChatGPT’s broader, more general audience.
Secondly, the study finds the share of messages related to companionship or social-emotional issues is fairly small.
Only 1.9% of ChatGPT messages are on the topic of Relationships and Personal Reflection and 0.4% are related to Games and Role Play.
I think this percentage is much lower than generally expected…
Doing & Asking
“Asking, Doing, or Expressing” refers to a classification of user interactions with chatbots where “Asking” involves seeking information or advice to aid decision-making.
“Doing” entails requesting the model to perform tasks by generating outputs.
And “Expressing” encompasses interactions that are neither asking for information nor requesting task performance.
The data from OpenAI’s study on ChatGPT usage shows Asking as the dominant and fastest-growing category, rising from roughly equal footing with Doing in mid-2024 (around 46% each) to 51.6% by late June 2025.
While Doing dropped to 34.6% and Expressing grew modestly to 13.8% — suggesting users are increasingly treating AI as a thinking partner for decision support rather than a pure task executor.
What I find interesting is the fact that we are still in the “asking phase”…and an evolution needs to take place to the “doing phase”…
The doing phase is more technical and involved, and it is also more consequential…but there are a number of users especially making use of Claude for coding that has made this transition.
Chief Evangelist @ Kore.ai | I’m passionate about exploring the intersection of AI and language. Language Models, AI Agents, Agentic Apps, Dev Frameworks & Data-Driven Tools shaping tomorrow.
