AI use by country in mid-2026: what Microsoft's data means for where you measure AI visibility
Microsoft's June 2026 diffusion data puts global generative-AI use at 18.8% of working-age people. How to use it to rank countries, and where it stops helping.
The short answer
Microsoft's data for June 2026 shows 18.8% of people aged 15 to 64 used a generative-AI product, with the UAE and Singapore far ahead. Use it to decide which of your markets to measure first and how many prompts each deserves. Do not read it as search share: it covers all generative-AI use, measured from Microsoft's own telemetry.
Key takeaways
- In Microsoft's report for June 2026, 18.8% of people aged 15 to 64 worldwide used a generative-AI product.
- The same data puts the UAE at 73.3% and Singapore at 64.3%, against 28.8% for the Global North and 16.2% for the Global South.
- High adoption is a reason to measure AI visibility in a market early, weighted by the revenue you expect there.
- The figure counts any generative-AI use from adjusted Microsoft telemetry, so it says nothing direct about AI search or which engine people use.
- Pair it with search-side signals, such as where Google's AI Mode is available, before setting country budgets.
In this article
What the report measured
Microsoft's AI diffusion report, published on 21 September 2026, estimates the share of people aged 15 to 64 who used a generative-AI product in June 2026, country by country. The global figure was 18.8%, about one point higher than in the first quarter.1
The gap between richer and poorer economies remains wide. Microsoft puts the Global North at 28.8% and the Global South at 16.2% for June 2026.1 At the top, the UAE reached 73.3% and Singapore 64.3%.1
The movers matter too. South Korea showed the largest absolute gain in the quarter, 3.5 percentage points, and Saudi Arabia climbed from 30th to 25th place.1 Japan rose 2.2 points from 22.5% in the first quarter.1
Share of people aged 15–64 using generative AI, June 2026
How it is built, and what it cannot tell you
The estimate comes from aggregated, anonymised Microsoft telemetry, adjusted for operating system and device market share, internet penetration and population. Microsoft says plainly that no single metric is perfect. Three limits follow for anyone planning AI visibility work.
First, it measures any generative-AI use: writing, coding, image tools and chat, as well as search-like questions. A country can rank high because office workers draft emails with AI, while buyers there still shortlist suppliers in classic search. Second, it is built from one company's view of devices, then adjusted; markets where Microsoft's footprint is thin rest more on the adjustment. Third, it does not report which engine people use, so a market dominated by one assistant looks the same as a market split across several.
Our view: treat it as a ranking of where AI habits are forming, not as a measure of AI search share. It answers "where are people used to asking AI?" well. It does not answer "where do buyers in my category ask AI about suppliers?"
A second signal: AI for a specific task
Task-level surveys give a narrower and lower picture. The Reuters Institute's Digital News Report 2026 surveyed online news users in 48 markets (mid-January to February 2026, with some questions repolled in March). It found 10% used an AI chatbot for news each week.2
The two numbers measure different things, so they do not conflict. Diffusion counts any use in a month; the news figure counts one task, weekly. The lesson for planning is that general adoption overstates use for any single job, including researching suppliers. Your own prompt results are the only measure of your category.
How should the ranking change where you measure AI visibility?
Use the diffusion ranking as one input when deciding which markets to measure first and how many prompts each gets. Combine it with two others: the revenue you expect from each market, and whether the engines your buyers use operate there.
Google's AI Mode is a useful example of the third input. Google said in October 2025 that AI Mode was adding more than 35 new languages and would be available in over 200 countries and territories.3 Availability is not use, but a market without the feature cannot show it in answers.
A practical rule follows. Measure first in markets with high adoption and meaningful revenue. Hold markets with high adoption and little revenue for later. Where adoption is low but revenue is high, run a small prompt set to set a baseline and re-check it each quarter. Our guide to budgeting across countries sets out the prompt counts.
For the two leaders, our earlier playbooks go further: UAE travellers planning with AI and financial firms in Singapore. In each, the prompts are written in the languages buyers use and run per country, because answers differ by market.
What we would watch next quarter
Watch the movers more than the leaders. A market rising fast, as South Korea did this quarter, is where AI answers about your category may form before competitors notice. Re-check the ranking when Microsoft publishes the next report, and compare it with your own prompt results in the same markets. Microsoft says that report will widen the measure to newer AI tools and expects the share to rise in almost all economies, so treat a jump next quarter partly as a change of method.1
Our view: adoption data tells you where to look, and only measurement tells you what buyers see. In a Sigzen AI audit we run each prompt on three separate days across six engines, using official APIs where offered and a licensed AI-answer data provider otherwise, so a country's result rests on more than one run.
Sources
- Microsoft, Global AI Diffusion Q2 2026: people aged 15–64 using generative AI, June 2026 (Sep 2026)
- Reuters Institute, Digital News Report 2026: online news users in 48 markets, YouGov, mid-Jan–Feb 2026 with some questions repolled in Mar 2026 (Jun 2026)
- Google, AI Mode expands to more languages and locations (Oct 2025)
