# Anthropic reports modest shifts in Claude use and releases new usage data

[Canonical reading](https://escapemeridian.com/readings/anthropic-reports-modest-shifts-in-claude-use-and-releases-new-12433396)

Anthropic’s report, published March 27, 2025, examines Claude.ai usage after the launch of Claude 3.7 Sonnet. Anthropic reports modest increases in coding, education and science use, while the balance between augmentation—people working with the model—and automation—asking it to complete tasks—remained unchanged. It says the model’s “extended thinking” mode was used most often for technical tasks, and describes how interaction patterns varied by occupation: copywriters and editors showed more task iteration, while translators and interpreters showed more directive use. These are patterns in Claude.ai conversations, not measures of how work or employment changed across the economy. Anthropic also releases a bottom-up taxonomy of usage, intended to capture activities that predefined occupational task lists might miss. The report says its O\*NET-based analysis may not represent how people actually use Claude for some occupations. It concludes that measuring AI’s economic effects will need to develop as models and their uses change. The report does not establish what these usage patterns mean for workers or labor-market outcomes.

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

## Original source

Anthropic · Article

Original source date: Mar 27, 2025 (2025-03-27T00:00:00.000Z).

Date basis: Date stated on the publisher’s page.

Added to Escape Meridian: 2026-11-14T18:53:25.379Z. First observed: 2026-10-18T09:41:20.563Z. Retrieved: 2026-10-28T03:13:49.648Z.

- [Economic Index: Insights from Claude 3.7 Sonnet](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Limits

Text as retrieved on October 28, 2026; originally published March 27, 2025. The analysis covers anonymized Claude.ai Free and Pro conversations, not a representative measure of AI use across workers or workplaces. Anthropic notes that O\*NET descriptions may not accurately represent how Claude is used for some occupations. The report describes usage patterns; it does not establish their effects on employment or labor-market outcomes.


## Open question

How well do Claude.ai conversation patterns represent AI use across workers and workplaces, and what do they imply for employment or changes to work?


## Source-stated

Anthropic reports modest increases in Claude.ai use for coding, education and science after Claude 3.7 Sonnet launched.

> we find relatively modest increases in coding, education, and scientific use cases

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic says the balance between augmentation and automation did not change in its newer data.

> The balance of augmentation and automation has stayed relatively constant in the two months between our data samples (V1 and V2)

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic reports that extended thinking mode was used predominantly for technical and creative problem-solving tasks.

> Our analysis reveals that Claude 3.7 Sonnet's extended thinking mode is predominantly used in technical and creative problem-solving contexts.

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic says its O\*NET occupational descriptions may not accurately represent some Claude.ai use.

> Note that the O\*NET descriptions may not be optimally representative of what Claude is being used for

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic releases a bottom-up dataset of usage clusters alongside its O\*NET-based analysis.

> It consists of 630 granular clusters, with associated descriptions, prevalence metrics, and automation/augmentation breakdowns

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source and contributor context

Anthropic · Publisher

The publisher is identified here. No individual author has been inferred.


## Source media

Media retrieval status: unchecked.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Anthropic

Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.

- [Explore this context](/entities/anthropic)
- [Reference](https://www.anthropic.com/news)
- [Reference](https://www.anthropic.com/research/team/economics)

## Claude

Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.

- [Explore this context](/entities/claude)
- [Reference](https://claude.com/product/overview)
- [Reference](https://www.anthropic.com/news)
- [Reference](https://www.anthropic.com/research/team/economics)

## Editorial provenance

Version 1. Publication ID: 26edb196-a1c2-7ff8-7824-1d0a43e5e8d8.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T18:49:32.699Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge\_admitted\_by\_model.

Observation: 3f77b45b88fd379c3f88fe80c74d4f7d217b819978f11960bc631fc5f34daa9c. Source text hash: 86c12dce1140805239702f55e7e94e0e91bde4b51459c40432182efd4e469fa0.


## Revision and source-check history

### Version 1 — 2026-11-14T18:53:25.379Z

Anthropic’s report, published March 27, 2025, examines Claude.ai usage after the launch of Claude 3.7 Sonnet. Anthropic reports modest increases in coding, education and science use, while the balance between augmentation—people working with the model—and automation—asking it to complete tasks—remained unchanged. It says the model’s “extended thinking” mode was used most often for technical tasks, and describes how interaction patterns varied by occupation: copywriters and editors showed more task iteration, while translators and interpreters showed more directive use. These are patterns in Claude.ai conversations, not measures of how work or employment changed across the economy. Anthropic also releases a bottom-up taxonomy of usage, intended to capture activities that predefined occupational task lists might miss. The report says its O\*NET-based analysis may not represent how people actually use Claude for some occupations. It concludes that measuring AI’s economic effects will need to develop as models and their uses change. The report does not establish what these usage patterns mean for workers or labor-market outcomes.

## Original source

Anthropic · Article

Original source date: Mar 27, 2025 (2025-03-27T00:00:00.000Z).

Date basis: Date stated on the publisher’s page.

Added to Escape Meridian: 2026-11-14T18:53:25.379Z. First observed: 2026-10-18T09:41:20.563Z. Retrieved: 2026-10-28T03:13:49.648Z.

- [Economic Index: Insights from Claude 3.7 Sonnet](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Limits

Text as retrieved on October 28, 2026; originally published March 27, 2025. The analysis covers anonymized Claude.ai Free and Pro conversations, not a representative measure of AI use across workers or workplaces. Anthropic notes that O\*NET descriptions may not accurately represent how Claude is used for some occupations. The report describes usage patterns; it does not establish their effects on employment or labor-market outcomes.


## Open question

How well do Claude.ai conversation patterns represent AI use across workers and workplaces, and what do they imply for employment or changes to work?


## Source-stated

Anthropic reports modest increases in Claude.ai use for coding, education and science after Claude 3.7 Sonnet launched.

> we find relatively modest increases in coding, education, and scientific use cases

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic says the balance between augmentation and automation did not change in its newer data.

> The balance of augmentation and automation has stayed relatively constant in the two months between our data samples (V1 and V2)

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic reports that extended thinking mode was used predominantly for technical and creative problem-solving tasks.

> Our analysis reveals that Claude 3.7 Sonnet's extended thinking mode is predominantly used in technical and creative problem-solving contexts.

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic says its O\*NET occupational descriptions may not accurately represent some Claude.ai use.

> Note that the O\*NET descriptions may not be optimally representative of what Claude is being used for

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source-stated

Anthropic releases a bottom-up dataset of usage clusters alongside its O\*NET-based analysis.

> It consists of 630 granular clusters, with associated descriptions, prevalence metrics, and automation/augmentation breakdowns

- [Supporting source](https://www.anthropic.com/research/anthropic-economic-index-insights-from-claude-sonnet-3-7)

## Source and contributor context

Anthropic · Publisher

The publisher is identified here. No individual author has been inferred.


## Source media

Media retrieval status: unchecked.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Anthropic

Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.

- [Explore this context](/entities/anthropic)
- [Reference](https://www.anthropic.com/news)
- [Reference](https://www.anthropic.com/research/team/economics)

## Claude

Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.

- [Explore this context](/entities/claude)
- [Reference](https://claude.com/product/overview)
- [Reference](https://www.anthropic.com/news)
- [Reference](https://www.anthropic.com/research/team/economics)

## Editorial provenance

Version 1. Publication ID: 26edb196-a1c2-7ff8-7824-1d0a43e5e8d8.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T18:49:32.699Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge\_admitted\_by\_model.

Observation: 3f77b45b88fd379c3f88fe80c74d4f7d217b819978f11960bc631fc5f34daa9c. Source text hash: 86c12dce1140805239702f55e7e94e0e91bde4b51459c40432182efd4e469fa0.


### uncertain — 2026-11-14T18:49:32.699Z — anthropic/claude-opus-5.5

Publication: 26edb196-a1c2-7ff8-7824-1d0a43e5e8d8

attribution — pass: The reading is a single Anthropic research report. It consistently attributes the findings to Anthropic, using phrases such as 'Anthropic reports' and 'It says'.

> Today, we’re releasing our second research report from the Index

summary — uncertain: The summary is faithful overall, but several central terms are left unexplained and no definitions are supplied. O\*NET is never described as the U.S. Department of Labor's database of occupational tasks. 'Extended thinking' mode is not described as letting the model think longer on complex questions. 'Task iteration' and 'directive use' are also unexplained. A general reader may not follow these points. Smallest repair: add a brief plain-language gloss for each term, or tooltips.

> maps each conversation to one of 17,000 tasks in the U.S. Department of Labor’s O\*NET database

date — pass: The stated date, March 27, 2025, matches the UTC day of originalPublished. The source is a publisher page, so no archive disclosure is needed. The page was modified after publication, but the claims the reading relies on are part of the core report body. The reading uses no later material and no hindsight framing. The source's 'Dec ‘25 - Jan ‘25' typo is not repeated.

support — pass: Each claim matches a source passage: the modest increases, the stable augmentation/automation balance, extended thinking used for technical tasks, the O\*NET caveat and the 630-cluster taxonomy. The reading's caveats about representativeness and labor-market effects are framed as limits, not as assertions. The question follows from the evidence. The page was edited after publication, but the cited passages appear to be part of the original report.

> we find relatively modest increases in coding, education, and scientific use cases, and no change in the balance of augmentation and automation

media — pass: No media is selected and no verified media candidates were supplied, so nothing needs to be checked.

## Recorded references

Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter.

No public references recorded here yet.

[Reference data (JSON)](/readings/12433396bf7aa302526c710d4e3a3fb98df220470c633412f6871006feb0937a/references.json)

## Corrections and challenges

[Send a correction or an opposing source](https://escapemeridian.com/corrections/)

[JSON record](https://escapemeridian.com/readings/anthropic-reports-modest-shifts-in-claude-use-and-releases-new-12433396.json)
