How the World is Using ChatGPT
AI for DC
Hi
Welcome (back) to The Prompt. AI isn’t well understood, but we learn a lot in our work that can help. In this newsletter, we share some of those learnings with you. If you find them helpful, make sure you’re signed up for the next issue.
[Data] How the world is using ChatGPT
In partnership with Harvard economist David Deming, our Economic Research team has produced a NBER working paper on how consumers are using ChatGPT. Based on the largest sample to-date of actual AI use, 1.5 million conversations, derived from the world’s most widely used AI tool, the paper is our first real window into how AI is being democratized and how ChatGPT is creating economic value – including what economists call “consumer surplus.”
Consumer surplus helps explain why something like ChatGPT can have a big impact even if people don’t pay much — or in the case of the vast majority of our users, anything — for it: the usefulness may far exceed the price, generating a lot of surplus value. That value is not captured by traditional economic indicators, though top economists have found that AI’s unmeasured consumer surplus is around $100 billion per year.
Read the paper here, and register here to catch our Chief Economist Ronnie Chatterji and David Deming talk through the findings on the OpenAI Forum. But The Prompt team has our own favorite data points and second-order takeaways, including a couple of not-outrageous postulations, and we’re sharing them here.
>700 million
… people around the world are using ChatGPT every week, and most of that use is through our free plan. Between most ChatGPT use being free, its widely varied use cases (see below), and broad and still growing adoption, ChatGPT usage will be a key indicator of AI’s impacts on the economy.
New technologies help people do things they couldn’t previously do. That extends beyond just what people do with the tools themselves, to the knock-on effects of what else they can do because they’re using the tools.
ChatGPT is putting the ability to create personal benefit into the hands of all these people. But because it enhances their ability to think, create, discover and produce, the personal benefit could be as simple as more time to spend on more fulfilling work, increasing productivity – or it could be as game-changing as actual societal benefit, such as progress toward new cures for disease.
30%
… of consumer ChatGPT use is for work. The study’s sample is only of consumer messages, not messages on the enterprise side. Yet even among consumers, nearly one third of activity is for work. While previous technologies launched in the workplace (the factory, the office) have tended to stay there, AI is a technology that consumers are adopting for work faster than industry is. This may be the first powerful technology in history that consumers are able to access as quickly as the world’s leading companies.
2.5 billion
For context on the rest, that’s the average daily volume of ChatGPT messages (or prompts).
52%
… of users have names that are typically feminine (which is how we approached the study in order to maintain user privacy) – the closing of a gender gap that was as wide as 20% feminine names to 80% masculine names around ChatGPT’s launch. This excludes users with names not typically masculine or feminine.
4x
Growth in adoption in the lowest income countries compared to the highest income countries as of this past May – another demographic gap closing.
46%
… of all messages sent are by users aged 18-25, although older age groups are gaining share over time. The share of work-related messages by age cohort is lowest among this youngest cohort, at 23%, which tracks given that many younger people may be using these tools for more, and more varied, tasks.
>10%
… of all consumer messages are for tutoring or teaching.
4.3%
… of ChatGPT consumer user messages are about coding, a relatively small share of our consumer sample.
2,087
The number of detailed individual work tasks that were mapped to consumer messages. That’s 2,087 different ways people are using ChatGPT.
~6x
The increase in message volume from July 2024 to July 2025, due to improvements in the capabilities of the models, and users discovering new cases for existing capabilities over time.
42%
The share of work messages related to writing. Interestingly, most of this is to edit and critique text, not write from scratch. By comparison, 24% of work messages are for practical guidance, and 10% are for technical help.
~10%
…of the world’s population – that’s what over 700 million weekly active users amounts to.
[About] OpenAI Forum
Explore past and upcoming programming by and for our community of more than 30,000 AI experts and enthusiasts from across tech, science, medicine, education and government, among other fields.
8:00 PM – 9:00 PM EDT on Sep 16
[Disclosure]
Graphics created by Base Three using ChatGPT.