Artificial intelligence use worldwide: how many people actually use AI?

Artificial intelligence seems to be everywhere.

It appears in the news, corporate presentations, productivity discussions, universities, LinkedIn, company strategy plans and nearly every conversation about the future of work.

ChatGPT, Gemini, Claude, Copilot and dozens of other platforms have made artificial intelligence one of the most visible technology topics in the world.

But there is a much more interesting question:

How many people actually use artificial intelligence?

And, more importantly:

How many use it deeply enough to change the way they work?

The answer depends on what we mean by “using artificial intelligence.”

Encountering an AI-generated summary in Google is not the same as opening ChatGPT once a month. Asking an AI for the capital of a country is also not equivalent to using it every day to draft documents, paying for a professional subscription, connecting it to company information or delegating an entire process to agents.

Once we separate these behaviors, a much more interesting reality appears.

AI already has an enormous audience.

But there is still a considerable distance between being exposed to AI, using it, paying for it and truly integrating it into work.

First: there are approximately 8.3 billion of us

The full universe must be our starting point.

United Nations projections place the world population in 2026 at approximately 8.3 billion people.

That is our 100%.

From there, the universe of people who could use advanced digital technologies progressively narrows.

Of 8.3 billion people, around 6 billion use the Internet

The International Telecommunication Union, ITU, estimates that approximately 6 billion people used the Internet in 2025, equivalent to about 74% of the world population.

This also means that around 2.2 billion people remained offline.

Using the projected 2026 world population as the reference, those 6 billion represent approximately 72% of humanity.

This already provides an important perspective.

Before we even discuss artificial intelligence, approximately one in four people on the planet remains outside the connected digital universe.

The first funnel

The figures do not align perfectly because the sources use different reference dates and methodologies. These numbers should therefore be understood as orders of magnitude, not as a simultaneous census conducted on the same day.

How many of those 6 billion people use the Internet every day?

This is where the first interesting measurement problem appears.

There is currently no single, harmonized and sufficiently recent global figure that allows us to say precisely that “X people use the Internet every day.”

The ITU does state that in most countries, the majority of Internet users connect daily. Even in lower-income economies, more than half of users access the Internet every day and, in many countries, the percentage approaches 90%.

We can therefore say with reasonable confidence that several billion people use the Internet daily.

What we should not do is invent an exact worldwide number that international sources have not yet published in a harmonized form.

This methodological detail matters.

One common problem in discussions about technology adoption is presenting estimates as though they were census results.

Being connected does not mean using the Internet for work

The next leap is often overlooked.

There are around 6 billion Internet users.

But there are not 6 billion digital workers.

The global employed population is considerably smaller.

The International Labour Organization reported that 284 million workers represented 7.9% of the worldwide employed population in 2025. That relationship places global employment at roughly 3.6 billion people.

In other words:

less than half of the world population is employed.

And among those approximately 3.6 billion workers are farmers, machine operators, drivers, construction workers, salespeople, domestic workers, restaurant employees, technicians, informal workers and millions of people whose jobs do not happen primarily in front of a computer.

There is no sufficiently comparable global figure that tells us how many of those 3.6 billion people use the Internet every day as a work tool.

That absence of data also matters.

The discourse around artificial intelligence is extraordinarily concentrated on office work and, particularly, on what is known as knowledge work.

The world of work is much larger.

Then we reach artificial intelligence

Measurement becomes even more complicated here.

At least four concepts are often mixed together:

These are different things.

Invisible AI: billions already use it without opening ChatGPT

Google provides a good example.

In June 2026, Google reported that AI Overviews had reached more than 2.5 billion monthly users within Search.

That represents approximately 30% of the world population.

But it would be incorrect to say that 2.5 billion people have become active users of generative artificial intelligence.

Many of them are simply using Google.

AI appears inside a product they were already using.

This distinction is fundamental:

exposure to AI is not the same as conscious AI adoption.

ChatGPT now exceeds 1 billion weekly users

In September 2026, OpenAI reports that its products reach more than 1 billion weekly active users.

To put that figure in context:

That is extraordinarily large.

It would therefore no longer be accurate to claim that generative artificial intelligence is used only by a small technological elite.

It is not.

But this does not mean that one in eight people on the planet has deeply integrated AI into their work.

Weekly user is a reach metric, not a sophistication metric.

A person who asks for a recipe and another person who has several agents analyzing corporate documents both count as users.

Gemini also exceeds 1 billion monthly users

Google announced in August 2026 that the Gemini app had surpassed 1 billion monthly active users.

But there is an important warning.

We cannot add:

1 billion ChatGPT users

and conclude that more than 2 billion different people use AI.

The same person may use ChatGPT, Gemini and Claude during the same week.

The populations overlap.

There is therefore no reliable public count of unique human beings who use any generative AI platform worldwide.

Any total that simply adds the user figures declared by different platforms would artificially inflate adoption.

A fundamental difference: trying AI is not adopting it

A global study conducted by the University of Melbourne in collaboration with KPMG surveyed more than 48,000 people in 47 countries.

Sixty-six percent said they intentionally used some form of AI with a degree of regularity.

The figure appears enormous.

But an important qualification is necessary: the research used online panels, and the Stanford AI Index itself notes that in some emerging countries the sample was more urban and educated than the general population, which may artificially increase the observed level of AI exposure.

This is an excellent example of why we must read beyond the headline.

“66% use AI” does not mean that 66% of the world’s 8.3 billion people use ChatGPT every week.

It means that 66% of participants in a large international online survey said they intentionally used AI technologies with some regularity.

Those are different statements.

AI at work: 58% have used it, but only 31% use it weekly or daily

The same study found that 58% of surveyed workers had intentionally used AI at work.

But when we raise the frequency threshold, the number falls.

Only 31% said they used it weekly or daily.

This difference is much more revealing.

We can think about three levels:

Contact: “I have used AI for work.”

Frequent adoption: “I use it every week.”

Operational integration: “It is part of my usual processes and performs parts of my work.”

Each level contains fewer people than the one before it.

Even in the United States, daily use of AI at work remains a minority behavior

Gallup offers an excellent contrast because it specifically measures American workers and distinguishes between frequencies of use.

In May 2026:

That is a considerable difference.

A statement such as:

“More than half of workers already use AI”

may be true under a particular definition.

But another statement can be true at the same time:

“Approximately 85% still do not use AI every day at work.”

The apparent contradiction disappears once we consider frequency.

Using AI every day does not mean work has been automated

This is probably the most important fact for understanding the real state of adoption.

Much of today’s AI use still resembles having an assistant, search engine, writer or adviser more than building truly automated work systems.

OpenAI’s data illustrates this difference.

OpenAI classifies interactions into three groups:

Asking: requesting information, an explanation or advice.

Doing: asking AI to produce something or perform a task.

Expressing: using it for conversation, opinions or personal expression.

Outside work, asking remains the dominant category.

The pattern changes in work contexts: users are more than twice as likely to ask ChatGPT to do or produce something as they are outside a professional setting.

That is precisely the transition we need to observe.

Not simply:

Do you use AI?

But rather:

What are you delegating to AI?

For many people, AI still functions as the next search engine

Earlier OpenAI research found that three broad categories represented approximately 77% of conversations in ChatGPT:

This is revealing.

Much of what people do with one of the world’s most advanced AI tools still feels quite familiar:

“Explain this to me.”

“Find information about this.”

“How do I do this?”

“Write this.”

“Correct this.”

“Summarize this.”

These are valuable uses.

But they are far removed from a company where AI:

That second scenario belongs to a much more advanced stage of adoption.

Agents are probably the true frontier

The Stanford AI Index 2026 offers another particularly revealing fact.

Although 88% of surveyed organizations use AI in some form and 70% use generative AI in at least one business function, the use of AI agents remained in the single digits across practically all business functions.

This may be one of the best measures of the difference between noise and transformation.

AI may be present in 88% of organizations.

But that does not mean 88% have rebuilt their operations around it.

Making ChatGPT available to employees is not the same as redesigning business processes.

Even among enterprise AI customers, there is an enormous depth gap

OpenAI’s own enterprise data shows an enormous difference between organizations that merely have access to AI and those that exploit it intensively.

OpenAI calls the 10% of organizations with the highest AI usage in its enterprise universe “frontier firms.”

Those companies currently generate 8.3 times more output tokens per active user than a typical company.

There is another, even more interesting signal.

Each week, Plugins are used by:

Plugins connect AI with other tools and business contexts.

Therefore, even inside organizations that have already purchased OpenAI technology, the use of advanced capabilities remains concentrated among a minority.

This may be a better definition of adoption:

not who has access to AI, but who has actually changed the way they work because of it.

The most interesting figure: only about 50 million people pay directly for ChatGPT

Here we find another important filter.

In 2026, OpenAI reported more than:

The platform later surpassed 1 billion weekly users.

Using those figures as a reference:

around 5% of weekly users would be paying directly for an individual subscription.

Viewed against the entire universe:

50 million represents approximately:

In other words:

fewer than one in every 100 people on the planet currently pays for an individual ChatGPT subscription.

This does not mean only those people generate revenue for OpenAI. There are enterprise plans, API usage, education and other commercial arrangements.

OpenAI also reported more than 9 million paid business users.

Some may overlap with people who also pay for individual accounts, so once again the figures should not be added indiscriminately.

How many people pay for Claude, Gemini and other AI tools?

Here we must acknowledge another data limitation.

There is currently no public, comparable census showing how many unique people pay for any generative artificial intelligence tool.

OpenAI does publish its number of individual subscribers.

Google reports subscription figures aggregated across multiple services and has disclosed millions of enterprise Gemini licenses, but it does not provide a directly comparable global number of consumers who pay specifically for Gemini.

Anthropic publishes enterprise metrics, revenue figures and Claude Code growth, but it also does not provide a comparable global number of individual Claude subscribers.

Therefore:

we cannot responsibly claim that exactly X million people worldwide pay for AI.

What we can say is that the known figure from the market leader is illustrative:

more than 1 billion people use ChatGPT every week, while around 50 million pay for an individual subscription.

The conversion from reach to payment remains relatively small.

The real artificial intelligence adoption funnel

A more useful way to think about AI is as a funnel.

Level 1: population

8.3 billion people

100% of the universe.

Level 2: connectivity

≈6 billion use the Internet

≈72% of humanity when the projected 2026 population is used as the denominator.

Level 3: AI exposure

Billions encounter AI embedded in search engines, phones, social networks, apps, financial services and work platforms.

Google AI Overviews, for example, already exceeds 2.5 billion monthly users.

Level 4: conscious use of generative AI

ChatGPT exceeds 1 billion weekly users.

Gemini exceeds 1 billion monthly users.

But there is substantial overlap between platforms.

Level 5: use at work

International studies show that a considerable share of workers have tried AI.

The figures fall when we require weekly use.

They fall again when we require daily use.

Level 6: payment

ChatGPT has more than 50 million individual subscribers.

That is approximately 5% of its weekly audience and less than 1% of global Internet users.

Level 7: process integration

Here we enter a much smaller group.

These are users and companies that connect AI to information, applications, tools and processes.

Level 8: truly agentic work

Here we find agents that can execute multistep processes and act on external systems.

Available evidence indicates that this level remains early and highly concentrated.

The mistake is to simply say “AI adoption is 60%”

There is no single AI adoption rate.

The right question should always be:

adoption of what?

People who have seen AI-generated content?

People who have used ChatGPT at least once?

Monthly users?

Weekly users?

Daily users?

Workers who use it?

Companies that bought a license?

Paying users?

People who automated tasks?

Organizations that integrated agents?

Each definition produces a radically different figure.

And all of them can be correct at the same time.

AI is massive as a phenomenon, but still early as work infrastructure

This is where the most interesting conclusion appears.

Generative artificial intelligence can no longer be described as a marginal technology.

More than 1 billion people use ChatGPT every week.

More than 1 billion use Gemini every month.

Billions receive AI-generated results inside products that are already part of their lives.

The stage of mass unfamiliarity has ended.

But we are much farther from a second stage:

operational transformation.

Most people do not pay directly for AI.

Most workers still do not use it every day.

Most companies have not rebuilt their processes around agents.

And even among organizations that have already purchased enterprise AI tools, there is a wide gap between average users and those who truly exploit advanced capabilities.

That is probably the most important story behind the numbers.

We are not at the end of the AI revolution. We are probably at the beginning of its second phase

The first phase was getting millions of people to try artificial intelligence.

That has already happened.

The second is turning a curiosity, a chatbot or an improved search engine into a genuine productivity tool.

The third will be turning it into infrastructure.

When a person stops thinking:

“I am going to open ChatGPT and ask it something”

and begins to think:

“this process happens automatically because of artificial intelligence,”

adoption changes in nature.

We move from using AI to working with AI.

And later to something different still:

having work performed by AI.

Current data indicates that the world has already made substantial progress through the first stage.

It is rapidly entering the second.

But it remains in the early phases of the third.

That is why artificial intelligence can appear omnipresent while still having enormous room to grow.

There is no contradiction.

AI has already won the battle for attention.

It is still fighting the battle for integration.

Key data

Methodological note

The figures above come from different studies conducted between 2025 and 2026 and should not be treated as a single simultaneous global census. The definitions of “user,” “active user,” “AI use,” “employee,” “weekly user” and “subscriber” are not equivalent.

When a comparable global figure does not exist, it is better to acknowledge the absence of data than to manufacture precision through extrapolation.