Anthropic Economic Index September 2025 Report

TL;DR

Anthropic’s September 2025 Economic Index shows that Claude’s usage is highest in wealthy nations and knowledge‑intensive US states, that businesses automate tasks far more than consumer users, and that directive automation has jumped from 27 % to 39 % of conversations in nine months, indicating growing trust in AI.

Overview of the September 2025 Economic Index

The third Anthropic Economic Index expands its scope to geographic data, compares usage over time, and adds anonymized API traffic from first‑party customers. The report quantifies how Claude is adopted across countries, US states, occupational categories, and interaction patterns (automation vs. augmentation).

Geographic Distribution of Claude Use

Global Share by Country

  • The United States accounts for 21.6 % of all Claude interactions, far ahead of the next largest user, India.
  • Brazil, Japan, and South Korea follow with similar shares.

Anthropic AI Usage Index (AUI)

  • AUI normalizes Claude use by each country’s share of the world’s working‑age population.
  • Countries with AUI > 1 use Claude more than expected; those with AUI < 1 use it less.
  • The top five AUI scores are Israel, Singapore, Australia, New Zealand, and South Korea.

Income Correlation

  • A 1 % increase in GDP per capita correlates with a 0.7 % higher AUI.
  • High‑income, technology‑oriented economies (e.g., Israel, Singapore) adopt Claude disproportionately relative to population size.
  • This mirrors historical patterns where general‑purpose technologies first amplify growth in richer regions, raising concerns about a new “great divergence.”

US State‑Level Adoption

Income Effect Within the United States

  • A 1 % rise in per‑capita GDP inside the US is linked to a 1.8 % higher population‑adjusted Claude use.
  • Income explains less variance than in cross‑country analysis, implying additional drivers.

Economic‑Structure Drivers

  • District of Columbia leads with an AUI of 3.82, dominated by document editing and information‑search tasks.
  • California shows a high share of coding‑related queries (third‑highest AUI overall).
  • New York features finance‑related tasks prominently (fourth‑highest AUI).
  • Hawaii demonstrates tourism‑focused usage at twice the national rate.

Evolution of Task Mix (December 2024 – September 2025)

Dominant Categories

  • Computer and mathematical tasks remain the largest share (≈ 37‑40 % of conversations).

Notable Shifts

Task Category Change (Dec 2024 → Sep 2025)
Educational instruction +40 % (9 % → 13 % of conversations)
Physical & social sciences +33 % (6 % → 8 %)
Management –40 % (5 % → 3 %)
Business & financial operations –50 % (6 % → 3 %)
  • Absolute conversation volumes grew across all categories, but the relative composition shifted toward knowledge‑intensive fields.

Income‑Driven Task Diversification

  • As per‑capita income rises, Claude use moves away from computer‑mathematical tasks and toward education, arts & design, office support, and the physical/social sciences.
  • The pattern is noisy but consistent across the dataset.

Interaction Patterns: Automation vs. Augmentation

Definitions

  • Automation – AI produces output with minimal user input.
    • Directive: user gives a single instruction, AI completes the task.
    • Feedback loop: user provides outcome feedback for model refinement.
  • Augmentation – User and AI collaborate.
    • Learning: user asks for information or explanations.
    • Task iteration: back‑and‑forth refinement.
    • Validation: user seeks AI feedback on their work.

Recent Trends

  • Directive automation rose from 27 % to 39 % of all Claude conversations.
  • Overall automation now accounts for 49.1 % of interactions, overtaking augmentation (47 %).
  • The shift suggests increasing user confidence in Claude’s ability to deliver high‑quality results on first attempts.

Cross‑Country Contrast

  • Higher per‑capita Claude use correlates with more augmentation and less automation.
  • A 1 % increase in population‑adjusted use predicts roughly a 3 % reduction in the automation share.
  • The causal mechanisms remain unclear; possibilities include early adopters’ willingness to automate or cultural/economic factors.

Business (API) vs. Consumer (Claude.ai) Usage

Task Distribution

Audience Computer/Mathematical Educational Arts & Entertainment
API customers 44 % 4 % 5 %
Claude.ai users 36 % 12 % 8 %
  • API traffic is more heavily weighted toward coding and administrative tasks; educational and creative tasks are under‑represented.

Automation Preference

  • API customers: 77 % of conversations exhibit automation (mostly directive); only 12 % show augmentation.
  • Claude.ai: Automation and augmentation are roughly balanced.
  • The disparity points to a potential labor‑market impact as businesses increasingly rely on AI for autonomous task execution.

Cost Sensitivity

  • Higher‑cost task categories (measured by average token consumption) see greater usage among API customers.
  • This positive correlation suggests that businesses prioritize tasks where AI’s value‑add outweighs token costs, rather than avoiding expensive interactions.

Implications and Outlook

  • Uneven Adoption – Wealthier nations and knowledge‑intensive US states adopt Claude more rapidly, raising the risk of a global productivity gap.
  • Shift Toward Automation – The rapid rise in directive automation indicates growing trust, but also foreshadows potential displacement in routine occupations.
  • Business‑Driven Automation – Companies are already using Claude autonomously at higher rates than individual consumers, hinting at early signs of AI‑driven workflow transformation.
  • Policy Relevance – The data provide early, empirical evidence for policymakers to consider targeted retraining, digital infrastructure investment, and equitable AI diffusion strategies.

Data Access and Further Exploration

  • The full dataset—including geographic breakdowns, task‑level patterns, and API usage metrics—is publicly available on the Anthropic Economic Index website and via the HuggingFace repository Anthropic/EconomicIndex.
  • An interactive dashboard lets users query Claude usage by US state, occupation, and country.

References

Sources

Related

  • Dispatch
    Anthropic Economic Index September 2025 report

    Anthropic’s September 2025 Economic Index report shows that AI adoption is highly uneven across geographies and enterprises, with usage concentrated in high‑income regions and dominated by coding and automation tasks.

  • Dispatch
    Anthropic Economic Index January 2026 Report – Economic Primitives

    Anthropic’s January 2026 Economic Index report introduces five new “economic primitives” that quantify how Claude is used, revealing geographic concentration, task complexity, skill bias, and productivity implications.

  • Dispatch
    Anthropic Economic Index March 2026 Report: Learning Curves

    Anthropic’s March 2026 Economic Index report shows that Claude usage diversified, lower‑wage tasks grew, and high‑tenure users achieve higher success rates, indicating learning‑by‑doing and potential skill‑biased impacts.

  • Dispatch
    Anthropic Economic Index January 2026 Report

    Anthropic introduces economic primitives to measure AI's impact on labor productivity, finding that Claude provides the greatest speedups for complex tasks requiring higher education.

  • Dispatch
    Anthropic Economic Index launch and first findings

    Anthropic announced the Economic Index, a data-driven study of AI usage across occupations based on millions of Claude.ai conversations, and released the underlying dataset for open research.