=== PAGE 1 ===
September, 2026
Global AI Diffusion
Q2 2026 Trends and Insights 


=== PAGE 2 ===
Global AI Diffusion Q2 2026 Trends and Insights 
Executive Summary
Global AI diffusion continued to climb in the second quarter 
of 2026. In June, AI usage was 18.8% of the working-age 
population worldwide, an increase of about 1% from the first 
quarter of this year. Growth was broad-based: almost every 
economy saw increased usage. 
Notably, the UAE (73.3%) and Singapore (64.3%) continue 
to top the leaderboard. South Korea again saw the largest 
absolute change, with an increase of 3.5 percentage points. 
Saudi Arabia climbed five places from 30th to 25th, the 
largest increase in rank. Japan recorded the largest relative 
change, growing 2.2 percentage points, or roughly 10% 
relative to its Q1 usage of 22.5%. Worldwide, the quarter 
brought continued widening of the AI gap between the 
Global North and South, with usage now at 28.8% in the 
North and 16.2% in the South. 
To track all these trends, we continue to measure AI diffusion 
as the share of people worldwide between ages of 15 
and 64 who have used a generative AI product during the 
reported period. This measure is derived from aggregated 
and anonymized Microsoft telemetry data and is adjusted 
to reflect differences in operating system and device-market 
share, internet penetration, and population. Additional 
details on the methodology are available in our AI Diffusion 
technical paper. [1]
AI diffusion by economy, June 2026
AI user share
10%
20%
30%
Insufficient data
UAE
73.3%
Singapore
64.3%
Ireland
49.9%
France
49.6%
Norway
49.4%
2

=== PAGE 3 ===
Global AI Diffusion Q2 2026 Trends and Insights
No single metric is perfect, and this one is no exception. 
Through the Microsoft AI Economy Institute, we continue 
to refine how we measure AI diffusion globally, including 
how adoption varies across countries in ways that 
best advance priorities such as scientific discovery and 
productivity gains. For this report, we rely on the strongest 
cross-country measure available today. Over the past 
year, there has been a significant rise in new AI tools and 
models. For our next report we will expand our measure to 
include usage of these new tools and expect this update 
to increase AI User Share across almost all economies, with 
larger projected increases in China.
In addition to our comparison of AI diffusion between the 
Global North and the Global South, we also find important 
differences in how AI is being used. Using Consumer 
Copilot usage insights, we examined these differences. The 
most notable difference is a higher share of education and 
learning use in the Global South. 
This report also includes a substantive discussion of open-
weight AI models. Open-weight AI model use is increasingly 
visible through application programming interface (API) 
usage of AI, where open models have captured a growing 
share of token volume. This shift could especially benefit 
the Global South by lowering access costs and enabling 
adaptation to local languages and needs, although 
infrastructure, connectivity, skills, and other foundational 
barriers remain.
Leading Economies
The United Arab Emirates (73.3%) and Singapore (64.3%) 
again lead the world by a wide margin, followed by Ireland 
(49.9%) and France (49.6%). Within the top 25, South Korea 
was again the standout riser, gaining 3.5 percentage points, 
the largest quarterly increase in the group, and climbing 
four places from 16th to 12th. South Korea has now posted 
the largest gain of any major economy for three consecutive 
reporting periods from H2 2025 to Q1 2026 to Q2 2026, 
adding nearly 15 points to its AI user share over the past 12 
months. Saudi Arabia saw the largest rise in rank, climbing 
five places from 30th to 25th and entering the top 25.
3

=== PAGE 4 ===
Global AI Diffusion Q2 2026 Trends and Insights 
Al diffusion over time by economy
Q1 2026 AI diffusion
Q2 2026 AI diffusion
1
United Arab Emirates
70.1%
2
Singapore
63.4%
3
Norway
48.6%
4
Ireland
48.4%
5
France
47.8%
6
Spain
44.2%
7
New Zealand
43.0%
8
United Kingdom
42.2%
9
Netherlands
42.1%
10
Qatar
41.8%
11
Australia
39.5%
12
Belgium
39.0%
13
Israel
38.1%
14
Switzerland
37.8%
15
Canada
37.3%
16
South Korea
37.1%
17
Sweden
36.1%
18
Austria
34.1%
19
Hungary
32.2%
20
T aiwan
31.8%
21
United States
31.3%
22
Denmark
31.2%
23
Germany
31.1%
24
Poland
31.0%
25
Italy
30.2%
26
Czech Republic
30.1%
27
Jordan
29.7%
28
Bulgaria
29.7%
29
Finland
29.5%
30
Saudi Arabia
29.4%
1
United Arab Emirates
73.3%
3.2pp
2
Singapore
64.3%
0.9pp
3
Ireland
49.9%
1.5pp
4
France
49.6%
1.8pp
5
Norway
49.4%
0.8pp
6
Spain
45.1%
0.9pp
7
New Zealand
44.1%
1.1pp
8
United Kingdom
43.7%
1.5pp
9
Netherlands
43.5%
1.4pp
10
Qatar
43.5%
1.7pp
11
Australia
40.7%
1.2pp
12
South Korea
40.6%
3.5pp
13
Belgium
40.3%
1.3pp
14
Switzerland
39.1%
1.3pp
15
Israel
38.7%
0.6pp
16
Canada
38.5%
1.2pp
17
Sweden
37.3%
1.2pp
18
Austria
35.4%
1.3pp
19
T aiwan
33.8%
2.0pp
20
Hungary
33.3%
1.1pp
21
United States
33.0%
1.7pp
22
Denmark
32.4%
1.2pp
23
Germany
32.3%
1.2pp
24
Poland
32.3%
1.3pp
25
Saudi Arabia
31.4%
2.0pp
26
Italy
31.4%
1.2pp
27
Czech Republic
31.2%
1.1pp
28
Jordan
31.2%
1.5pp
29
Bulgaria
30.7%
1.0pp
30
Finland
30.6%
1.1pp
4

=== PAGE 5 ===
Global North
Global South
-40%
-20%
0%
20%
40%
Searching for information
Create/edit text or code
Feedback or advice
Shopping/product research
Self-directed learning
Create/edit images
Technical support
Schoolwork/academic support
-14.1%
+6.5%
-10.3%
+22.4%
-27.5%
+12.6%
-36.9%
+16.9%
-13.4%
+29.2%
-14.9%
+32.3%
-8.3%
+3.8%
-14.9%
+32.3%
Global AI Diffusion Q2 2026 Trends and Insights
The Global North-South divide 
The underlying gap between the Global North and 
Global South continued to widen in Q2 2026. In the 
second quarter of 2026, 28.8% of the population in the 
Global North used generative AI, up from 27.5% in Q1 
2026, a gain of 1.3 percentage points. In the Global 
South, usage reached 16.2%, up from 15.4%, a gain of 
0.8 percentage points. 
Underlying this gap in adoption are important differences in 
how AI is being used. Learning stands out in particular in the 
Global South. Our analysis of a subset of consumer Copilot 
conversations uses a privacy-preserving classifier to identify 
the broad “intent” of a conversation – the primary task in 
which Copilot users engage. [2] We grouped AI usage by 
economy into the Global North and South, as above, and 
compared the share of conversations for each intent with 
its share globally.
In the Global South, self-directed learning and skill 
development represent a 29.2 percent larger share of 
conversations than they do globally, and schoolwork 
and academic support accounts for a 32.3 percent 
Region
H2 2025  
AI Diffusion
Q1 2026 
AI Diffusion
Q2 2026  
AI Diffusion Change
Global North 24.7% 27.5% 28.8% +1.3
Global South 14.1% 15.4% 16.2% +0.8
World 16.3% 17.8% 18.8% +1.0
Gap between Global North  
and Global South
10.6pp 12.1pp 12.6pp
Global North and South intent prevalence relative to the global mix
Intents ordered by global prevalence; North is 68.5% of traffic and South is 31.5%
5

=== PAGE 6 ===
Global AI Diffusion Q2 2026 Trends and Insights 
larger share. Importantly, the Global South has a much 
younger population than the North. Therefore, it is likely 
that demographic differences are at least partially driving 
these results. 
The remaining intents tilt in more familiar directions. The 
Global North leans toward consumer uses such as shopping, 
search, and feedback. And in the Global South, image, text, 
and code creation and editing take an outsized share. Taken 
together, these results suggest that learning support may be 
a particularly important entry point to AI in regions that are 
earlier in their adoption curve. This is consistent with recent 
studies of Bing Copilot, Gemini, and Claude, which also find 
schooling or coursework to be a larger part of observed AI 
use in lower-income countries. [3][4][5]
However, we suggest cautious interpretation of these 
findings. These results apply only to the early adopters using 
AI in these regions, and may not extrapolate as AI usage 
expands to the broader population. Additionally, the data do 
not explain why learning features more prominently in the 
Global South. Even so, the pattern adds a dimension to the 
North-South gap discussed above: in regions where diffusion 
still lags, a larger share of existing use is devoted to learning.
Open-weight models are broadening AI Diffusion
Many AI models are “open” in the sense that they release 
their trained weights, though the training process and 
code may or may not be open. These open-weight 
models can be downloaded, modified, and run by other 
organizations, giving developers much greater freedom to 
deploy and adapt them than closed models accessed only 
through a proprietary service. Their growing importance 
was reflected in a July 2026 open letter signed by 
Microsoft, NVIDIA, Meta, IBM, Hugging Face, the Linux 
Foundation, and other companies and organizations, 
which argued that open weights can make advanced 
AI more accessible and adaptable while allowing 
organizations to match “the right model to the right job 
at the right cost.” [6]
Open-weight models also highlight an important 
dimension of AI diffusion that is difficult to capture 
through consumer-facing measures alone. Currently, 
our primary AI diffusion measure in this report estimates 
the share of the working-age population that has used 
a generative AI product. But open-weight models are 
especially likely to be encountered indirectly: through 
an application, coding agent, or other service that 
accesses a model programmatically through an API, 
rather than through a branded chatbot (though some 
chatbots do build on open models). Some organizations 
also download and run open weights on their own 
infrastructure, where their usage may not appear in 
any centralized data source. This makes the API market 
a useful complementary window into diffusion. One 
of the best public sources of data on open-weight 
model usage is OpenRouter, which provides a common 
interface through which developers can access hundreds 
of open and closed models and route requests among 
them. [7] Its position between users and many different 
inference providers allows an unusual view of which 
models developers are choosing in practice, although its 
developer-oriented user base should not be interpreted 
as representative of all AI users.
This market is becoming increasingly important because 
the capability gap between open-weight and closed 
models has narrowed rapidly. Research comparing 
successive generations of models across independent 
6

=== PAGE 7 ===
Sources: Microsoft Al Economy Institute; Artificial Analysis.
Open-weight models catch up to the closed frontier within a few months
Artificial Analysis Intelligence Index by model release date
Closed weights
Open weights
Individual models
0
10
20
30
40
50
60
February
March
April
May
June
July
August
September
53.4
44.9
Global AI Diffusion Q2 2026 Trends and Insights
benchmarks shows a recurring pattern: new closed models 
push the capability frontier forward, but leading open-
weight models often reach similar performance within a 
matter of months. [8] Recent releases suggest that even 
this distinction is becoming increasingly task-specific. 
Moonshot AI’s Kimi K3, whose full weights were released 
in July, still trails the strongest proprietary models on some 
broad evaluations, including in Moonshot’s own technical 
report. [9] But on other measures it is already competitive 
with—or ahead of—leading closed models. As of September 
14, for example, Kimi K3 Max ranked in the top 5 on 
Arena’s WebDev coding leaderboard, ahead of a number of 
proprietary alternatives at a fraction of the cost. [10] These 
rankings should not be interpreted as showing that Kimi K3 
is universally superior, but they illustrate how quickly open-
weight models have moved from lower-cost alternatives to 
competitors at parts of the capability frontier.
That improvement in capability is increasingly translating 
into usage. OpenRouter’s public analysis of more than 
100 trillion tokens found that open-weight models 
had grown to roughly one-third of token volume by 
late 2025, with major releases such as DeepSeek, Kimi, 
Qwen, and GPT-OSS producing increases that persisted 
beyond their initial launch periods—evidence that 
adoption reflected production use rather than only 
experimentation. [11] The shift accelerated further in 2026. 
More recently our own analysis of public OpenRouter 
data shows that open-weight models now represent 75% 
of OpenRouter token usage. These figures are not global 
market shares: OpenRouter users are unusually likely 
to experiment across providers, while direct usage of 
closed frontier models as well as locally deployed open 
models are missing from the data entirely. Nevertheless, 
the magnitude and direction of the change provide a 
7

=== PAGE 8 ===
Sources: Microsoft Al Economy Institute; Artificial Analysis. 
Near-frontier intelligence, far lower cost per task
Artificial Analysis Intelligence Index vs average cost per benchmark task
Global AI Diffusion Q2 2026 Trends and Insights 
clear signal that open-weight models are becoming an 
increasingly important part of the API-based AI economy. 
This pattern is also consistent with emerging research 
using more granular OpenRouter data, which finds a 
broad movement toward greater use of open models as 
their capabilities improve. [12]
The economics of that shift could become increasingly 
important as AI moves from experimentation to large-
scale deployment. Analyses show that the capability gap 
between open and closed models is not only decreasing, 
but is decreasing at a more rapid pace. [12] Further, the 
most recent public data from Artificial Analysis shows that 
some open-weight models deliver similar capabilities at 
substantially lower cost per task. For example, the recently 
released GLM 5.3 Flash has similar performance to GPT 5.6 
Sol at ’high’ reasoning level at roughly one third the cost. 
The Sol models from OpenAI were, as of just a few weeks 
ago, near the frontier for intelligence. At the same time, 
businesses are becoming more attentive to the cost of 
rapidly growing AI workloads. The once-popular practice 
sometimes described as “tokenmaxxing”—treating 
rising token consumption as a measure of successful AI 
adoption—can become expensive as agents and other 
applications make repeated model calls. Companies are 
therefore increasingly routing different tasks to different 
models rather than automatically using the most capable 
frontier system for every request; Reuters reported open-
weight models gaining ground as businesses scrutinized 
AI spending and searched for less costly alternatives. [13] 
The emerging principle is not simply to minimize token 
use or always choose the cheapest model, but to match 
the model to the task: reserving expensive frontier 
intelligence for problems where its additional capability 
creates value while using lower-cost models for the much 
larger set of workloads where they perform sufficiently 
well. This is the same economic logic emphasized in the 
recent open-weight letter.
Closed weights
Open weights
Cost-performance frontier
0
15
30
45
60
$0
$2
$4
$6
$8
Artificial Analysis Intelligence Index
Cost per task
GLM-5.3-Flash
GPT-5.6 Sol high
Fable 5.1 max
Astra max
Astra low
GLM-5.3 max
Kimi K3 max
GPT-5.6 Sol max
Grok 4.6 high
GPT-5.6 Luna max
8

=== PAGE 9 ===
Sources: Microsoft Al Economy Institute analysis of Hugging Face.
Developer activity around Al applications has expanded rapidly
Cumulative surviving projects hosted on Hugging Face Spaces
0
500K
1M
1.5M
Mar '22
Sep '22
Mar '23
Sep '23
Mar '24
Sep '24
Mar '25
Sep '25
Mar '26
Aug '26
First AI diffusion report (Oct 2025)
797K Spaces
August 2026
1.46M Spaces
1.8× since first report
Global AI Diffusion Q2 2026 Trends and Insights
These economics could be particularly consequential 
for the Global South. As discussed earlier in this report, 
AI adoption reached 28.8 percent of the working-age 
population in the Global North in the second quarter of 
2026 compared with 16.2 percent in the Global South, 
where gaps in electricity, connectivity, digital skills, and 
local-language support continue to constrain diffusion. 
Open-weight models cannot eliminate these foundational 
barriers. But they can reduce two additional ones: the 
cost of accessing capable AI and the difficulty of adapting 
models to local needs. Because weights can be modified, 
non-profits, businesses, governments, and universities can 
fine-tune or otherwise adapt existing models rather than 
train frontier systems from scratch. For example, companies 
have fine-tuned open-weight models to increase their 
expertise in specific domains like law. [14] Further, 
research from Microsoft’s AI for Good Lab illustrates the 
potential of this approach: its Bring Your Own Language 
framework improved performance for Chichewa and 
Māori by approximately 12 percent on average across a 
set of benchmarks by refining data and adapting existing 
language models, while also developing a pathway for 
extremely low-resource languages such as Inuktitut. [15] 
Open weights make this type of localization easier and 
allow models to be deployed through competing API 
providers—or, where infrastructure permits, locally—while 
giving users greater control over how they are customized. 
If capable open-weight models continue to trail the frontier 
by months rather than years while remaining available at 
a fraction of the inference price, they could become an 
important pathway for extending increasingly capable AI to 
populations and economies that have so far benefited least 
from AI’s diffusion.
Growth in AI Tools
Alongside the growth of open-weight models, 
developer activity around AI applications has expanded 
rapidly. Hugging Face Spaces, which hosts runnable 
AI applications and demos ranging from personal 
experiments and research projects to more production-
oriented applications, grew from approximately 124,000 
Spaces at the end of 2023 to 1.46 million by September 
2026. Since our first AI Diffusion report in November 
2025, the total has increased by approximately 83%. 
Although Spaces should not be interpreted as distinct 
9

=== PAGE 10 ===
Global AI Diffusion Q2 2026 Trends and Insights 
commercial AI products, their growth provides evidence 
of broadening experimentation and development at the 
application layer.
This growth in AI applications also has implications for 
how AI diffusion is measured. Many tools introduced since 
our first report have already attracted significant usage 
but are not included in our current measure. To address 
this gap, we plan to expand the list of AI tools covered 
in our next report, providing a more comprehensive 
view of AI usage worldwide. Initial analysis suggests that 
broader coverage will raise measured diffusion across 
nearly all countries, with the largest effect in China, where 
several widely used domestic tools are currently not 
included. These increases will primarily reflect improved 
measurement of existing usage rather than a sudden 
change in underlying adoption.
10

=== PAGE 11 ===
Global AI Diffusion Q2 2026 Trends and Insights
Citation and data availability
This report is based on the following technical paper:
Misra, A., Wang, J., McCullers, S., White, K., & Lavista Ferres, 
J. (2025). Measuring AI Diffusion: A Population-Normalized 
Metric for Tracking Global AI Usage.  
arXiv. https://doi.org/10.48550/arXiv.2511.02781
Also available at: https://aka.ms/AI_Diffusion_Technical_Report 
The underlying AI Diffusion data used in this report 
is publicly available:
AI Diffusion dataset (Q2 2026 update):
https://github.com/microsoft/ai-diffusion-report/blob/
main/data/AI_Diffusion_Q22026_Update.csv
When using this data, please cite the technical paper above 
and include a link to the dataset to ensure reproducibility 
and version transparency.
11

=== PAGE 12 ===
Global AI Diffusion Q2 2026 Trends and Insights 
[1] A. Misra, J. Wang, S. McCullers, K. White, and J. 
L. Ferres, “Measuring AI Diffusion: A Population-
Normalized Metric for Tracking Global AI Usage,” 
Nov. 4, 2025, arXiv:2511.02781. doi: 10.48550/
arXiv.2511.02781. https://arxiv.org/abs/2511.02781
[2] B. Costa-Gomes, S. Chen, C. Hsueh, D. Morgan, P. 
Schoenegger, Y. Shah, S. Way, Y. Zhu, T. Adeline, M. 
Bhaskar, M. Suleyman, S. Spielman, It’s About Time: The 
Copilot Usage Report 2025: The Temporal and Modal 
Dynamics of Copilot Usage, Dec. 2025. Microsoft AI. 
https://microsoft.ai/wp-content/uploads/2025/12/
What_people_do_with_Copilot-8.pdf
[3] M. I. G. Daepp and I. Slaughter, “How Early Adopters 
Used Generative AI Worldwide: Variation by Country 
Income and Language,” May 2026, arXiv:2605.30685. 
doi: 10.48550/arXiv.2605.30685. https://arxiv.org/
pdf/2605.30685
[4] Z. Iscenko, S. Strand, Y. Chen, G. Aimard, M. Codreanu, 
V. Sampathkumar, et al., “Google’s AI & Economy 
ATLAS v1.0: Mapping Gemini Usage in the Economy,” 
Jul. 2026. https://ai.google/static/documents/
GoogleATLASv1.pdf
[5] R. Appel, M. Massenkoff, P. McCrory, M. McCain, R. 
Heller, T. Neylon, and A. Tamkin, “Anthropic Economic 
Index report: Economic primitives,” Anthropic, Jan. 
15, 2026. https://www.anthropic.com/research/
anthropic-economic-index-january-2026-report
[6] Microsoft, “Open Weights and American AI 
Leadership,” 2026. https://www.microsoft.com/en-us/
corporate-responsibility/topics/open-weight/
[7] OpenRouter, “OpenRouter,” [Online]. Available: 
https://openrouter.ai/
[8] Artificial Analysis, “Artificial Analysis AI Trends,” 2026. 
[Online]. Available: https://artificialanalysis.ai/trends
[9] Kimi Team et al., “Kimi K3: Open Frontier Intelligence,” 
Jul. 2026, arXiv:2607.24653. https://arxiv.org/
abs/2607.24653
[10] Arena.AI, “Code Arena | WebDev Leaderboard,” 2026. 
[Online]. Available: https://arena.ai/leaderboard/
code/webdev
[11] M. Aubakirova, A. Atallah, C. Clark, J. Summerville, and 
A. Midha, “State of AI: An Empirical 100 Trillion Token 
Study with OpenRouter,” 2026, arXiv:2601.10088. 
https://arxiv.org/abs/2601.10088
[12] F. Nagle and D. Yue, “The Latent Role of Open Models 
in the AI Economy,” 2025. https://papers.ssrn.com/
sol3/Delivery.cfm?abstractid=5767103
[13] A. Soni, “American AI Model Makers Smell an 
Opportunity,” Reuters, Aug. 12, 2026. https://
www.reuters.com/technology/artificial-
intelligence/american-ai-model-makers-smell-an-
opportunity-2026-08-12/
[14] C. Qi, V. Rengarajan, J. Pereyra, N. Grupen, and G. 
Pereyra, “Harvey Tenet Research Preview,” Harvey, 
Aug. 20, 2026. https://www.harvey.ai/blog/post-
training-update-harvey-tenet
[15] S. W. Zamir, W. Hamidouche, B. B. Amor, L. Marotti, I. 
Becker-Reshef, and J. L. Ferres, “BYOL: Bring Your Own 
Language Into LLMs,” 2026, arXiv:2601.10804. https://
arxiv.org/abs/2601.10804
References
12

=== PAGE 13 ===
Global AI Diffusion Q2 2026 Trends and Insights
Economy Q1 2026 Q2 2026 Q2 Change
United Arab Emirates 70.1% 73.3% 3.2
Singapore 63.4% 64.3% 0.9
Ireland 48.4% 49.9% 1.5
France 47.8% 49.6% 1.8
Norway 48.6% 49.4% 0.8
Spain 44.2% 45.1% 0.9
New Zealand 43.0% 44.1% 1.1
United Kingdom 42.2% 43.7% 1.5
Netherlands 42.1% 43.5% 1.4
Qatar 41.8% 43.5% 1.7
Australia 39.5% 40.7% 1.2
South Korea 37.1% 40.6% 3.5
Belgium 39.0% 40.3% 1.3
Switzerland 37.8% 39.1% 1.3
Israel 38.1% 38.7% 0.6
Canada 37.3% 38.5% 1.2
Sweden 36.1% 37.3% 1.2
Austria 34.1% 35.4% 1.3
Taiwan 31.8% 33.8% 2.0
Hungary 32.2% 33.3% 1.1
United States 31.3% 33.0% 1.7
Denmark 31.2% 32.4% 1.2
Germany 31.1% 32.3% 1.2
Poland 31.0% 32.3% 1.3
Saudi Arabia 29.4% 31.4% 2.0
Italy 30.2% 31.4% 1.2
Czech Republic 30.1% 31.2% 1.1
Jordan 29.7% 31.2% 1.5
Bulgaria 29.7% 30.7% 1.0
Finland 29.5% 30.6% 1.1
Slovenia 29.0% 30.1% 1.1
Costa Rica 28.5% 29.7% 1.2
Vietnam 26.5% 27.9% 1.4
Lebanon 27.3% 27.7% 0.4
Oman 26.5% 27.6% 1.1
Portugal 26.4% 27.4% 1.0
Croatia 26.1% 27.2% 1.1
Slovak Republic 26.1% 27.1% 1.0
Colombia 24.5% 25.9% 1.4
Dominican Republic 24.8% 25.8% 1.0
AI Diffusion Data Source
13

=== PAGE 14 ===
Global AI Diffusion Q2 2026 Trends and Insights 
Economy Q1 2026 Q2 2026 Q2 Change
Uruguay 24.6% 25.7% 1.1
Serbia 24.1% 25.3% 1.2
Lithuania 24.3% 25.1% 0.8
Jamaica 24.0% 25.0% 1.0
Japan 22.5% 24.7% 2.2
Panama 23.3% 24.2% 0.9
Chile 22.7% 24.0% 1.3
South Africa 23.1% 23.9% 0.8
Argentina 21.9% 23.4% 1.5
Bosnia And Herzegovina 22.1% 23.2% 1.1
Malaysia 21.8% 22.6% 0.8
Kuwait 21.1% 22.6% 1.5
Georgia 20.5% 21.6% 1.1
Greece 20.8% 21.5% 0.7
Mexico 20.1% 21.4% 1.3
Philippines 20.1% 20.7% 0.6
Ecuador 19.5% 20.3% 0.8
Brazil 19.1% 20.2% 1.1
El Salvador 18.3% 20.1% 1.8
Albania 18.5% 19.2% 0.7
Moldova 18.5% 19.1% 0.6
Turkey 17.4% 18.9% 1.5
Azerbaijan 17.7% 18.7% 1.0
India 17.6% 18.5% 0.9
Romania 17.5% 18.0% 0.5
Mongolia 16.7% 17.9% 1.2
China 16.4% 17.5% 1.1
Peru 16.4% 17.3% 0.9
Guatemala 16.4% 17.3% 0.9
Kazakhstan 15.9% 16.9% 1.0
Gabon 15.0% 15.7% 0.7
Libya 15.0% 15.5% 0.5
Egypt 14.8% 15.5% 0.7
Namibia 15.1% 15.4% 0.3
Botswana 14.8% 15.0% 0.2
Indonesia 14.1% 14.8% 0.7
Nepal 14.2% 14.6% 0.4
Honduras 14.0% 14.5% 0.5
Senegal 13.9% 14.4% 0.5
Tunisia 13.5% 13.9% 0.4
14

=== PAGE 15 ===
Global AI Diffusion Q2 2026 Trends and Insights
Economy Q1 2026 Q2 2026 Q2 Change
Algeria 13.3% 13.8% 0.5
Cote D'Ivoire 13.1% 13.7% 0.6
Thailand 12.4% 13.4% 1.0
Bolivia 12.7% 13.3% 0.6
Zambia 13.1% 13.3% 0.2
Iraq 12.5% 13.0% 0.5
Iran 12.6% 12.8% 0.2
Paraguay 12.2% 12.6% 0.4
Nicaragua 11.8% 12.5% 0.7
Morocco 11.7% 12.0% 0.3
Pakistan 11.4% 11.9% 0.5
Gambia 11.5% 11.5% 0.0
Mozambique 10.9% 11.4% 0.5
Angola 10.9% 11.4% 0.5
Madagascar 10.9% 11.4% 0.5
Malawi 10.9% 11.4% 0.5
Guyana 10.3% 11.3% 1.0
Suriname 10.3% 11.3% 1.0
Venezuela 10.3% 11.3% 1.0
French Guiana 10.3% 11.3% 1.0
Myanmar 10.0% 10.4% 0.4
Togo 10.1% 10.4% 0.3
Guinea 10.1% 10.4% 0.3
Niger 10.1% 10.4% 0.3
Liberia 10.1% 10.4% 0.3
Nigeria 10.1% 10.4% 0.3
Benin 10.1% 10.4% 0.3
Mali 10.1% 10.4% 0.3
Ghana 10.1% 10.4% 0.3
Guinea-Bissau 10.1% 10.4% 0.3
Sierra Leone 10.1% 10.4% 0.3
Mauritania 10.1% 10.4% 0.3
Burkina Faso 10.1% 10.4% 0.3
Belarus 9.6% 10.3% 0.7
Kyrgyzstan 9.5% 10.0% 0.5
Russia 9.5% 9.9% 0.4
Lesotho 9.8% 9.9% 0.1
Ukraine 9.5% 9.5% 0.0
Chad 8.7% 9.1% 0.4
Central African Republic 8.7% 9.1% 0.4
15

=== PAGE 16 ===
Global AI Diffusion Q2 2026 Trends and Insights 
Economy Q1 2026 Q2 2026 Q2 Change
Cameroon 8.7% 9.1% 0.4
Democratic Republic Of The Congo 8.7% 9.1% 0.4
Congo 8.7% 9.1% 0.4
Haiti 8.5% 8.9% 0.4
Kenya 8.7% 8.9% 0.2
Zimbabwe 8.5% 8.9% 0.4
Laos 7.8% 8.4% 0.6
Bangladesh 7.8% 8.0% 0.2
Uganda 7.6% 8.0% 0.4
Ethiopia 7.6% 8.0% 0.4
Somalia 7.6% 8.0% 0.4
Eritrea 7.6% 8.0% 0.4
South Sudan 7.6% 8.0% 0.4
Sudan 7.6% 8.0% 0.4
Burundi 7.6% 8.0% 0.4
Tanzania 7.6% 8.0% 0.4
Armenia 7.4% 7.7% 0.3
Papua New Guinea 7.7% 7.6% -0.1
Sri Lanka 7.3% 7.6% 0.3
Rwanda 7.2% 7.6% 0.4
Uzbekistan 7.2% 7.6% 0.4
Syria 7.5% 7.4% -0.1
Cuba 6.7% 6.8% 0.1
Tajikistan 6.1% 6.3% 0.2
Afghanistan 6.1% 6.3% 0.2
Turkmenistan 6.1% 6.3% 0.2
Cambodia 5.8% 6.1% 0.3
16

=== PAGE 17 ===