AI statistics 2026

How many people use AI?

Most people now. 64% of US adults use AI, up from 61% in 2025, and 25% of Americans use it every day, up from 19% (Menlo Ventures).

That's across a survey of 5,067 US adults, run with Morning Consult in July 2026.

Worldwide, generative AI reached 53% of people within three years, faster than the personal computer or the internet (Stanford HAI).

How many people use ChatGPT?

1.2 billion people a week. That's the number OpenAI gave at its developer conference on September 29, 2026 (OpenAI).

In the US, 60% of people who use AI use ChatGPT, up from 45% in 2025 (Menlo Ventures).

Which AI assistants do people use most?

ChatGPT and Gemini, with Claude far behind but growing fastest (Menlo Ventures).

ChatGPT and Gemini lead among US AI users (%)

1ChatGPT

60%

2Gemini

58%

3Claude

20%

Menlo Ventures, The State of Consumer AI 2026 · 5,067 US adults, July 2026

Claude nearly tripled its share in a year, from 7%. The average AI user now uses 3.0 assistants, up from 2.2.

The Gemini app has 950 million monthly users (Alphabet)

Google's models also process 22 billion tokens a minute through its API, and nearly 90% of the Fortune 100 use Gemini Enterprise.

How many people pay for AI?

More than half of users. 55% of AI users pay for at least one AI product, roughly 90 million Americans (Menlo Ventures).

A few heavy users pay for most of it. The 14% of payers who spend $100 or more a month make up 60% of consumer AI spending.

How many businesses use AI?

About 1 in 4 in the US. {{bt\_use}} of businesses used AI in the two weeks from {{bt\_ref}}, up from {{bt\_first}} in the two weeks to {{bt\_first\_ref}} (US Census Bureau).

US businesses that used AI in the past two weeks, 2026 (%)

1Two weeks to Sep 6

23.8%

2Two weeks to Aug 9

22.4%

3Two weeks to Jul 12

21.5%

4Two weeks to Jun 14

20.6%

5Two weeks to May 3

19.5%

6Two weeks to Apr 5

19.8%

7Two weeks to Mar 8

18.9%

8Two weeks to Feb 8

17.5%

9Two weeks to Jan 11

18.2%

US Census Bureau, Business Trends and Outlook Survey · about 200,000 US employer businesses asked each round, every two weeks

More are on the way: {{bt\_plan}} expect to use AI in the next six months. The Census Bureau asks every two weeks, so this chart updates with each new round.

How much money goes into AI?

More than ever. Companies invested $581.7 billion in AI in 2025, up 130% in a year (Stanford HAI).

Stanford notes that private investment likely understates China's spending, since much of it runs through government funds.

What is the best AI model right now?

{{ai\_top1\_name}} from {{ai\_top1\_maker}}, with {{ai\_top1\_score}} on the Artificial Analysis Intelligence Index, which averages 10 evaluations of agents, coding, science and reasoning (Artificial Analysis).

The highest-scoring AI models (score)

1Claude Opus 5.5 (Anthropic)

57.6

2Claude Sonnet 5.5 (Anthropic)

56

3Claude Fable 5.1 (Anthropic)

53.4

4GPT-6 Astra (OpenAI)

52.7

5Claude Opus 5 (Anthropic)

50.8

Artificial Analysis Intelligence Index, via OpenRouter · 94 models with a score, checked September 29, 2026

Models got much better at real work this year. AI agents' success on real-world tasks rose from 20% in 2025 to 77.3%, and cybersecurity agents now solve 93% of problems, up from 15% in 2024 (Stanford HAI).

Compare every model's score and price in my AI model comparison, updated every week.

How many AI models are there?

Hundreds. OpenRouter, which lets developers use most AI models through one account, lists {{ai\_models}} models from {{ai\_makers}} makers (OpenRouter).

Open models are close behind. The best open-weight model, {{ai\_open1\_name}} from {{ai\_open1\_maker}}, scores {{ai\_open1\_score}}.

How much does AI cost to use?

Less than a dollar per million tokens, for the median model. That's about 750,000 words for {{ai\_median\_price}} (OpenRouter).

The top models cost more: {{ai\_top1\_name}} is {{ai\_top1\_price}} per million tokens. But {{ai\_u1\_name}} scores {{ai\_u1\_score}} for {{ai\_u1\_price}}. See the cheapest and the best value in the AI model pricing board.

How fast are developers building with AI?

Faster every month. GitHub counted {{gh\_ytd}} new public repositories this year to {{gh\_until}}, against {{gh\_last}} by the same date last year, up {{gh\_growth}} (GitHub).

New GitHub repositories this year, by AI topic (repositories)

1llm

103,068

2ai-agents

92,291

3claude-code

75,571

4mcp

71,880

5rag

37,204

6ai-agent

31,344

7mcp-server

24,454

8openai

23,045

GitHub search · public repositories created this year, to September 28, 2026

GitHub's own report counted 4.3 million AI repositories in 2025, and 1.1 million public ones that call an AI model, 693,867 of them created in a single year (GitHub Octoverse).

Is AI taking jobs?

It's starting with young workers. Employment among software developers aged 22 to 25 has fallen nearly 20% since 2024, even as older developers' headcount grows (Stanford HAI).

The same pattern shows up in other jobs with high AI exposure, like customer service, and executives expect more cuts ahead.

How much energy does AI use?

A lot, and more each year. AI data center power capacity reached 29.6 gigawatts, about what New York State uses at peak demand (Stanford HAI).

Training one frontier model, Grok 4, produced an estimated 72,816 tons of CO2 equivalent, about what 17,000 cars emit in a year.

How to use these numbers

Before you pick an AI tool or plan around AI:

• Compare models on score and price, not on name
• Try a cheaper model first: many score close to the top for a fraction of the price
• Expect your customers and competitors to use AI already
• Check whether AI crawlers and agents can read your site
• Watch the business adoption numbers every few weeks: they move fast

Then choose the model that fits the job, and check its score again next month.

Key takeaways

Here's what the AI statistics say:

• AI is mainstream – ChatGPT reaches 1.2 billion people a week, and 64% of US adults use AI.
• Businesses are catching up – 23.8% of US businesses used AI in the latest two weeks, up from 17.7% at the start of the year.
• People pay for it – 55% of AI users pay for at least one AI product.
• The best models keep changing – Claude Opus 5.5 leads today, and 196 models were added this year.
• Good AI is cheap – the median model costs $0.80 per million tokens.
• Developers are all in – new GitHub repositories are up 79% on last year.

For more, see my AI model comparison and email marketing statistics, and try my free ChatGPT prompt generator.