CLASS READERS RESEARCH · July 22, 2026

AI & Learning in 2026: What 2 Million Conversations Reveal

We analyzed the latest open dataset from Anthropic's Economic Index (May 2026, released June 26) to answer one question: how big a role does learning actually play in how the world uses AI?

13.2%
of all Claude conversations worldwide are about Education & Learning — one of the largest single categories of AI use
16.5%
of usage is coursework — studying, homework and school-related tasks
1 in 4
AI conversations in much of the developing world is about education

Learning is not a niche use of AI. It's a main one.

Globally, Education & Learning accounts for 13.2% of everything people ask AI to do — spanning math tutoring, grammar instruction, language learning, reading comprehension and more. Add the "learning" collaboration pattern (where the human's goal is to understand, not to delegate) and you get 16.6% of all human-AI interactions.

In emerging markets, education dominates AI use

The most striking pattern is geographic. In many countries, education isn't just a use of AI — it's the use:

Algeria
32.6%
Zambia
28.8%
Tunisia
28.2%
Peru
27.2%
Iraq
26.5%
Indonesia
21.9%
India
19.3%
Australia
16.1%
United Kingdom
15.1%
United States
9.3%

Education & Learning as a share of each country's total Claude usage, May 2026. Gold = emerging markets, blue = selected high-income markets.

Teachers are AI's quiet power users

When conversations are mapped to occupations, educational instruction ranks among the top occupational profiles in dozens of countries — reaching 16–20% of all usage in places like Indonesia, Pakistan, Bangladesh, Ghana and Peru. Where classroom resources are scarce, teachers are turning to AI to prepare lessons, explain concepts and create materials.

In the countries where English is most eagerly learned, education is the single biggest thing people do with AI — and teachers lead the way.

Where AI adoption runs hottest

Per working-age capita, the heaviest AI-using countries are Australia (6.4× its population share), Singapore (5.8×), Switzerland (5.0×), New Zealand (4.8×) and Canada (4.1×) — English-speaking or English-fluent markets where parents are already comfortable putting AI-era tools in front of their kids.

What this means for early readers

Two worlds are converging: high-income families use AI-native tools by choice, and emerging-market families use them out of necessity — often specifically to learn English. Both need the same thing for young children: structured, level-appropriate reading practice, not a chat box. That's the gap read-along storybooks fill: every word highlighted as it's narrated, twelve levels that grow with the reader, quizzes and recording to make practice active.

Try Class Readers free
For journalists & bloggers: all figures above come from Anthropic's openly licensed Economic Index dataset (May 2026 release) and were independently aggregated by Class Readers. We're happy to share the analysis notebook, country-level tables or custom charts — email hello@classreaders.com.

Method

Source: Anthropic Economic Index, aei_claude_ai_2026-06-26.csv, May 2026 period (claude.ai Free/Pro/Max conversations, privacy-preserving aggregation by Anthropic). Education share = "Education & Learning" node of the request-topic hierarchy (pct metric) per geography. Adoption = usage_per_capita_index (usage share ÷ working-age population share). Occupation shares = SOC "Educational Instruction and Library" major group. Rounded to one decimal.