Digital Life · Humanity estimated
How many people are using AI right now?
How many people are using AI right now? See a transparent ChatGPT-anchored scenario with weekly-user conversion, assumed session time and limitations.
people, approximately
0.053% of the world population in this model
Model sensitivity (not a confidence interval): 660K–8.14M
≈ Modelled estimate · updated continuously
Typical estimate · modelled
Typical 24 Hours
The interactive 24-hour curve loads with the live model.
This curve shows a typical modelled day, not historical observations.
How this using ai estimate works
user pool × active days/week × minutes/day ÷ 7 ÷ 1,440 × local allocation
Anchor the scenario to 700 million weekly ChatGPT consumer users reported in the September 2025 research. Convert an assumed three active days × sixteen minutes into 48 minutes per weekly user, then divide by seven days. Allocate the audience with country Internet-population weights and local day/evening timing. Equal weekday weights avoid pretending that a verified global workweek pattern exists.
What the number means
How many people are using AI right now? See a transparent ChatGPT-anchored scenario with weekly-user conversion, assumed session time and limitations. Device access, platform reach, age and simultaneous media use vary widely between regions. The number is a broad modelled estimate, not direct tracking of individuals.
This estimate is for curiosity and perspective. Activities can overlap, and the totals should not be added together as if every category were exclusive.
Assumptions and limitations
The anchor is historical and covers one product, not all AI. Geographic penetration and service availability are not measured by the allocation proxy. The site does not observe prompts or active sessions. Country populations and civil clocks are source-based; country behaviour is not observed live. Survey values are transferred only as explicit proxies. The displayed range is editorial sensitivity, not a statistical confidence interval.
Source scope and reference years
- OpenAI / NBER — How people are using ChatGPT (2025): 15 September: 700 million weekly active users; consumer plans. Does not measure all AI products or session minutes.
Direct answers
Questions people ask
Can weekly active users be displayed as people online now?
No. The model assumes three active days per week and sixteen minutes per active day. Multiplying 700 million weekly users by 3 × 16/7 and dividing by 1440 yields the daily-average scenario.
Does this include every AI assistant?
No. The reviewed source is ChatGPT consumer usage. The number is a proxy scenario, and audiences from other assistants cannot be added without measuring overlap.
Are sixteen minutes and three days research findings?
No. They are explicit modelling assumptions. The source supports the 700-million weekly audience, not those session values.
What changes this estimate?
A published weekly audience supplies a starting point. How many days each person uses an assistant and how long they stay active are separate assumptions; rapid product growth can make the fixed anchor stale.
This is a ChatGPT-anchored proxy for consumer AI-assistant sessions, not a census of all AI products. Search algorithms, recommendation systems, embedded background AI and automated API traffic are outside scope.
Baseline sources
The anchor is historical and covers one product, not all AI. Geographic penetration and service availability are not measured by the allocation proxy. The site does not observe prompts or active sessions. Country populations and civil clocks are source-based; country behaviour is not observed live. Survey values are transferred only as explicit proxies. The displayed range is editorial sensitivity, not a statistical confidence interval.
- United Nations — World Population Prospects 2024
- UN WPP 2024, original demographic workbook (GEN/F01/Rev.1)
- IANA Time Zone Database — civil time zones
- OpenAI / NBER — How people are using ChatGPT
- Unicode CLDR — regional weekend conventions
Status: Source-informed simulation · explicit assumptions · Data vintage: Source years: 2025; reviewed 2026-09-08. Population and clock review: 8 September 2026. Activity-specific limitations are stated above.