Hi everyone! 1.2 billion users every week. That's exactly the audience OpenAI cited for ChatGPT at the end of September at its developer conference. I have to admit, I was impressed by that figure, and I understand it can raise questions like: am I falling behind? And who are these 1.2 billion? It seems everyone around me knows how to do things I haven't even started on, or maybe do differently. And even if I use some chat myself, and some model, and other AI tools, am I using them the way everyone else does?
Just recently I had two conversations like this in my own life. I was telling an acquaintance from my professional circle about something, and in the course of the conversation I mentioned the company Anthropic by name several times. The company that makes Claude. And in reply I heard: "What is Anthropic?" Another time, a young man from my circle, when I asked which model he uses, or maybe which chat he uses, asked me: "What do you mean?" And at the same time, these 1.2 billion ChatGPT users got me thinking. Who are these 1.2 billion, really? Where are they?
You watch tech videos, and people are already setting up agents, automating, setting up entire workflows, creating AI-native companies, and meanwhile someone nearby doesn't even know the name of the service. Of course, these two conversations don't really describe the world at all, and they don't describe reality. But in exactly the same way — let's be honest — YouTube videos don't describe the world either, and don't describe reality. There are about 8.3 billion people on Earth, and if you simply compare the numbers, you get roughly one in seven. That's a huge scale.
But that doesn't at all mean that every seventh person on the planet actively uses ChatGPT. Let's imagine two users. One asked ChatGPT to write a birthday greeting for a friend. The other goes through documents every day, automates processes, writes code. And both of them count in the weekly audience. But obviously they're mastering and using the technology in completely different, unequal ways. So today I want to figure out, first of all, what happens after they open the chat. What exactly are these 1.2 billion doing?
Where AI is used most often, and what of all this really changes our work and changes our lives. The first discovery is that mass-market AI looks nothing like it does in the tech feed. OpenAI has its Signals project, with open statistics on ChatGPT usage.
In the latest update, the August one, the available information is valid as of June 2026. So, the researchers examined a sample of 300,000 ChatGPT messages. It's very important to understand that this isn't 300,000 unique people; first and foremost, it's 300,000 messages. And another very important point: we're talking specifically about personal accounts. This analysis excludes Codex, and it also excludes all business products. So, what are the three largest categories among these messages?
They're recognizable, and I think we've already shared this information on our channel more than once, from one angle or another. Practical advice accounts for about 32% of messages, working with text — 22%, information seeking — 19%. Together, these three categories make up literally 73%. So, what does practical advice mean? Practical advice — that means: explain this topic to me, help me prepare for a conversation. How do I work through this task? Working with text covers all sorts of different things, from translating a text to actually composing it, to generating it.
Information seeking ranges from some specific fact to, well, choosing a product. And here I got a bit of a sense of a mismatch, I guess, between the news feed and the actual data. Because if you watch new product demos, if you hear the figure 1.2 billion, it seems like the whole world is already building digital employees, digital companies, digital — I don't know — homes, digital relationships. Nevertheless, in reality the bulk of ChatGPT messages are about very familiar needs: to figure something out, to formulate, to translate, to find, to choose.
Of course, it's very important to understand here — and I don't mean to say that people use AI in a primitive way. Of course not. And any request, even the tiniest, shortest, simplest one, can become the start of a very big analysis. And in general, even a letter can matter enormously. What I'd rather emphasize, I suppose, is that this mass adoption doesn't equal mass automation. And interestingly, by the way, the remaining 25% of messages aren't all agents either. The remaining 25% of messages include conversations about oneself, conversations about communication, about relationships, and they include images and media.
That perhaps the main mass-scale change, the main impact of AI, isn't that 1.2 billion people have, so to speak, learned to build agents, but that millions of people have gained a completely new way of dealing with very straightforward everyday tasks. And maybe not necessarily the most complex ones, but ones that are accessible every day. Let's move on. What happens next?
And the person got an explanation, came to understand the topic better, checked the report and made a better decision. Of course, the charts and the study's data don't show any of this. The second discovery — a conclusion I more or less came to for myself after getting to know this study from OpenAI's Signals project — is that we are, after all, shifting somewhat, moving from generating text to working through situations. That is, there's a year-over-year trend, and you can see that the share of practical advice grew from 28% to 32%, while working with text fell from 26% to 22%.
Information seeking barely changed. Here, perhaps — once again — we're talking about hypotheses and conclusions that I'm drawing; they're strictly within my own frame of reference and my own understanding. I'm curious what you think and how you see it. Share in the comments. I think it's extremely important right now, more than ever, to exchange opinions and share our reasoning, so that we can keep broadening our perception and understanding of AI and how it's applied. So, perhaps for some people the chat is becoming not just a "make me a finished text" kind of tool, but, more and more, a habitual way of working through a situation as it arises.
That is, before, when a question came up, we'd open a search engine, write to a colleague, gather information, and now there's this step: "let me first work through something, work through it with AI." And if that's really happening, then of course the consequences are much broader than what we see in demos of new features. Because, once again, it's not only the tool that's changing, and not only the tool that's developing — our habitual way of solving problems, of starting work on tasks, is changing too.
I think this is also one of the recurring themes we touch on here on our channel. When I heard 1.2 billion users, I was really curious what geography we were talking about and where these 1.2 billion users are.
Does geography turn the familiar picture upside down? Let's look at the map. In OpenAI's Signals ranking by number of ChatGPT messages per inhabitant in the second quarter of 2026, Singapore is in first place, Malta is second, then come the Netherlands and Montenegro. The US, by the way, is in 51st place. And that really surprised me at first. But let's dig into it. The United States — that's one of the main centers of AI development, and yet in terms of messages per inhabitant in this particular product it's, yes, far from first.
Here, of course, precision matters a lot again, because the ranking doesn't test Americans' knowledge, doesn't take corporate AI into account, and doesn't measure a whole range of other things — other ways of using AI at all, and other AI tools. It measures personal ChatGPT messages relative to population. Anthropic, in its May data — admittedly, I couldn't find anything more recent — offers a different perspective. There they look at usage of Claude and Cowork, and about 20% of global usage comes from the United States, 7% from India, and under 1% from Singapore.
Well, here it seems that a small country doesn't have to account for a large share of the global volume. But the researchers compared these shares. What share of global Claude usage comes from a country, and what share of the world's working-age population lives there. And for Singapore, the first share came out almost six times higher. So we have two different sources, and both Anthropic — Anthropic's study — and OpenAI's study show a very similar signal. That Singapore stands out significantly in activity relative to its population.
Of course, there's also another question here, right — not how many messages people send, but what share of people use AI at all. And on that point, I found an interesting study from Microsoft. This study covered various AI services, and according to the September update, with data for June 2026, about 19% of people aged 15 to 64 worldwide use AI today. In the UAE that figure is 73%, in Singapore 64%, while in the United States and Poland, for example, it's around 33% and 32%.
So now, as a result, we see three quite different measurements: message intensity, relative country activity, and prevalence of use among the population.
Obviously, to compare countries with one another, you need to use all of these factors, you need to use all of these components — and they mustn't be mixed. And of course, they show how unevenly AI is entering people's lives. It's interesting, of course, why Singapore and, say, the United Arab Emirates end up so high, and the ranking itself gives no answer to that question at all. You can look at the conditions, and I found a few interesting examples of how the adoption, development and promotion of AI is supported in these countries.
In Singapore, for example, the largest bank, UOB, together with a government agency, helps small companies choose AI tasks and build various prototypes. In the United Arab Emirates, AI is being integrated into key government digital services in partnership with tech companies. I mean, of course, this isn't proof of why these countries have such a high rate of using AI tools at all. But still, I think these examples show that adoption — that's not only each individual person's decision to download some app and ask questions like, well, how to peel tomatoes.
Of course, things like whether a person has access to the tool and some clear rules also matter a great deal, and a high ranking at the level of a given country, obviously isn't proof that people in that country are smarter and more talented, and a low one isn't evidence of a lack of interest — it's all much broader. The next thing, I suppose, is a conclusion that's important to keep in mind, one we're now coming to, namely,
that the 1.2 billion OpenAI talked about at its developer conference at the end of September is, of course, far from the whole world of AI. Very often, including on our channel, we talk specifically about OpenAI with ChatGPT and Anthropic with Claude. But that audience is certainly not the same as all the people who use AI. There's Gemini, there's DeepSeek, there's Qwen, there are specialized products, various corporate solutions, AI features inside all sorts of different work programs and apps.
And a person can use a great many different tools and services and never open, say, a personal ChatGPT at all. China is very telling in this respect. At the end of September, the China Internet Network Information Center reported that by mid-2026 China had more than 700 million active users of generative AI. Of course, China has its own big ecosystem, including DeepSeek and Qwen. And obviously, on the one hand, you're tempted to just add 700 million to 1.2 billion and get effectively two billion people, but of course that's not how it works.
How a user is defined here, and the possible overlaps — they differ. These are completely different methodologies, but even so, the point somehow doesn't get any weaker for it. ChatGPT — that's a huge part of the market today, but it's not the whole market. And if we talk about the world's use of AI only through, say, one product, then of course we only see — we're really, really narrowing our view, really narrowing our current understanding. Not to mention that AI has already stopped being just a separate chat, right — it's showing up everywhere around us: in search, in email, in documents, in work systems.
And the question of how many people use AI is getting harder and harder to answer. And some of that usage, I think, will soon simply become impossible to count. And another part of usage will become, as it were, invisible to us — invisible to the user themselves, in fact. The fifth discovery — well, not a discovery, it's more of a conclusion for me, one I found interesting to learn about. Who are these users? What age, what group might they even be? It could be young people, or people who are maybe older.
And interestingly, again according to the same Signals study, in June 2026 roughly two thirds of messages came from people aged 18 to 34. What's more — and this is interesting — the share of messages from people 35 and older grew from 28% to 34%. I mean, really, it's clear again here that we're analyzing the composition of messages rather than the reach of an age group, and we don't know why the share grew: whether new users came in or existing ones became more active. But this idea, which I also sometimes hear among ordinary people, that AI — that's a tool only for the very young, I think that kind of explanation looks very simplistic, and it's eroding more and more, especially given our earlier point that the use of AI is becoming so embedded that it's already turning fairly invisible.
By the way, there's movement in the age of AI users, too. The 65-plus group isn't as active: there it went from 1.9% to 2.5%. But even so, given the numbers we're talking about, that's a fairly serious, large scale.
So what overall picture do we end up with? 1.2 billion — that's simply a huge weekly reach for one product. And the bulk of messages in ChatGPT — that's practical advice, texts and information seeking. The geography is very uneven, and the share of messages from people 35 and older is growing. And ChatGPT — this is very important to understand — is, of course, part of a much broader ecosystem. But, you know, along with all this, I'm constantly trying to answer one question. And it's actually why I dive into this kind of analysis at all: how much better have people started to work and live thanks to this?
Because thanks to this analysis we know for sure that a message was sent, but we don't know how that answer was used, or how a person's life or environment may have changed. Whether that decision turned out better. And even within a single category, the complexity and the consequences are completely different. I'm coming back, essentially, to the question I started with. To the fact that this FOMO, this fear that you've already missed something, the fear that you're not where those 1.2 billion are, it can be very acute these days, it's being heavily fueled by all the mass media, by all the content the world around us is creating today.
But after all this analysis, I'd like to give, as it were, neither a reassuring answer nor some frightening conclusion. Because you can't say that the 1.2 billion can do things I can't do, that someone else can't do, that you can't do. We don't know what these people can do, we don't know what results they achieve. That's why I think it's very important not to try to judge yourself by the number of prompts, by age, by country, and not to invent some universal maturity test, but to keep asking yourself questions: which tasks do I now solve with AI that I solved differently before? What has actually gotten better?
Speed, quality, depth, access, perhaps, to some kind of information? And can I check that the improvement really happened? I think we've talked about these three starting questions many times, in different videos of ours, in different podcasts, on different topics. Because it's precisely this kind of attentive, mindful attitude that broadens our perception and the quality of how we use the tool that the world is, in fact, giving us today. And I'm very interested in your concrete examples: what you've started doing better with AI, and where, despite all of AI's promises, you maybe still can't get an effective workflow going.
Write them in the comments under this video — that would be really great, I'd love to read them. See you next time.