AI Use and Occupations is an independent research project by Fatih Kansoy. It studies how published records of generative-AI activity can be described across geographies and occupational classifications. The project combines a global evidence inventory with detailed audits of European classification routes and a native-classification benchmark for the United States.
The geographic design follows the data. The primary occupation evidence covers 121 geographies in May 2026. Europe is the principal classification case because the source O*NET categories must be related to ESCO, ISCO and UK SOC categories before they can be read beside European labour statistics. The United States supplies the benchmark in which source occupation codes can be joined to employment without that European crosswalk.
Independence and attribution
This project is not affiliated with or endorsed by Anthropic. Claude and the Anthropic Economic Index are the work of Anthropic. The transformations, interpretations and methodological proposals presented here are independent.
The sources include the Anthropic Economic Index; O*NET from the US Department of Labor, Employment and Training Administration; ESCO from the European Commission; and official statistical sources including ONS/Nomis, Eurostat, ILOSTAT, BLS and OECD PIAAC. Source attribution, vintages and transformations accompany the datasets. Inclusion does not imply endorsement by a source producer.
There is no blanket licence covering every source-derived file. Dataset-specific terms govern reuse, including required attribution and identification of adaptations. Restricted source vectors and respondent-level survey records are excluded from public downloads. For PIAAC aggregates, the required acknowledgement is: Programme for the International Assessment of Adult Competencies (PIAAC), Organisation for Economic Co-operation and Development (OECD), Paris.
Research materials
The site provides an occupation explorer, a geography inventory, a methods account, validation evidence, a full paper, a technical data note and a dataset catalogue. The primary occupation observations cover April and May 2026. Earlier one-week task observations from August 2025, November 2025 and February 2026 remain separate diagnostics because their windows and classification procedures are not directly comparable with the later monthly series.
The site distinguishes the observation period, source release and project build. A recent site update does not imply equally recent usage observations. The analysis uses a frozen source snapshot so that a displayed result can be traced to its evidence.
Scientific status
The results describe classified, published consumer conversations. They do not identify users' jobs, represent the workforce, measure time saved or estimate causal effects on employment, pay or productivity.
The independent public-input proxy for Anthropic's exposure score falls short of its prespecified numerical gate; exact reconstruction remains blocked by unavailable original inputs. European mapped values are semantic descriptors rather than allocated national usage shares. Reporting grades summarize publication and employment support under explicit post-review rules; they are not validated grades of representativeness or data quality. These limits are part of the reported result.
Corrections
Reproducible corrections are welcome. A useful report identifies the geography or occupation, observation period, measure, source and evidence supporting the proposed change. Contact and broader research information are available at fatih.ai.
Citation
Kansoy, Fatih (2026). AI Use and Occupations. Independent research dataset and website. fatih.ai/ai-use-and-occupations.
For a specific result, also record the observation period, dataset vintage, mapping route and shared-view or downloaded-file identifier. Cite Anthropic and the relevant statistical producer when reusing their underlying material. The full scholarly references are collected in the bibliography.
BibTeX
@misc{kansoy2026aiuseoccupations,
author = {Kansoy, Fatih},
title = {AI Use and Occupations},
year = {2026},
howpublished = {Independent research dataset and website},
url = {https://fatih.ai/ai-use-and-occupations/},
note = {Accessed YYYY-MM-DD}
}