Sources and licences

Every figure on PathMosaic comes from one of the sources below: the research first, then the data releases we use, each with its licence. Where a licence asks for credit, the credit is given in the wording it asks for.

Research on AI and work

The first five are the studies BLS combines into its exposure groups.

Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2024). GPTs are GPTs: Labor market impact potential of LLMs. Science, 384(6702), 1306–1308.

What AI could do.
Eisfeldt, A. L., Schubert, G., & Zhang, M. B. (2023, revised 2026). Generative AI and firm values. NBER Working Paper 31222.

What AI could do.
Felten, E., Raj, M., & Seamans, R. (2021). Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses. Strategic Management Journal, 42(12), 2195–2217.

What AI could do.
Tomlinson, K., Jaffe, S., Wang, W., Counts, S., & Suri, S. (2025). Working with AI: Measuring the applicability of generative AI to occupations. arXiv:2507.07935. Microsoft Research.

What people use AI for, from Copilot conversations.
Massenkoff, M., & McCrory, P. (2026, March 5). Labor market impacts of AI: A new measure and early evidence. Anthropic.

What people use AI for, from Claude conversations.
Machovec, C., Rieley, M. J., & Rolen, E. (2025, February). Incorporating AI impacts in BLS employment projections: Occupational case studies. Monthly Labor Review. U.S. Bureau of Labor Statistics.

How BLS adjusts its projections for AI.
Brynjolfsson, E., Chandar, B., & Chen, R. (2026, August 12 update). Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence. Stanford Digital Economy Lab.

Early-career employment in exposed occupations.

Research on careers

Rounds, J., Su, R., Lewis, P., & Rivkin, D. (2010, February). O*NET Interest Profiler Short Form psychometric characteristics: Summary. National Center for O*NET Development.

How reliable the interest scores are.
Hoff, K. A., Song, Q. C., Wee, C. J. M., Phan, W. M. J., & Rounds, J. (2020). Interest fit and job satisfaction: A systematic review and meta-analysis. Journal of Vocational Behavior, 123, 103503.

What interest fit does and doesn't predict.
Schubert, G., Stansbury, A., & Taska, B. Employer concentration and outside options. SSRN Working Paper 3599454.

The career moves data, used with the authors' permission on condition of citing this paper.

Data and licences

The release in use is named for each, so you can check a figure against the original.

U.S. Bureau of Labor Statistics U.S. Government work (public domain)

BLS 2025-35, 2026-08-27

AI exposure groups, projected growth, openings a year, median pay and typical education for each occupation.

Files: ai-exposure-categories.xlsx, nem-occupational-coverage.xlsx, nem-onet-to-soc-crosswalk.xlsx, occupation.xlsx
O*NET / U.S. Department of Labor CC BY 4.0 attribution required

O*NET 31.0

Occupation descriptions, tasks, work activities, subjects, education, Job Zones, interest ratings and related occupations.

Files: occupation_data.csv, job_titles.csv, sample_of_reported_titles.csv, task_statements.csv, task_ratings.csv, essential_skills.csv, transferable_skills.csv, knowledge.csv, software_skills.csv, related_occupations.csv, education.csv, education_categories.csv, training_and_experience.csv, training_and_experience_categories.csv, job_zones.csv, job_zone_reference.csv, career_interest_types.csv, content_model_reference.csv, gwas_to_iwas.csv, tasks_to_dwas.csv, gwas_to_iwas_to_dwas.csv

Credit, as the licence asks: This site includes information from the O*NET 31.0 Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. The makers of this site have modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications.
O*NET / U.S. Department of Labor CC BY-ND 4.0 (O*NET Career Exploration Tools) attribution required

O*NET Interest Profiler Short Form, paper-and-pencil version

The 60 activities on the student page and how they are scored.

Files: IPSF_PP.pdf

Credit, as the licence asks: This page includes information from the O*NET Career Exploration Tools by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY-ND 4.0 license.
Microsoft Research CC BY 4.0 attribution required

"Working with AI" v1.1

How often each work activity comes up in Copilot conversations.

Files: ai_applicability_scores.csv, soc_metrics.csv, iwa_metrics.csv, soc_to_iwas.csv, soc_iwa_weights.csv, physical_tasks.csv

Credit, as the licence asks: Kiran Tomlinson, Sonia Jaffe, Will Wang, Scott Counts, and Siddharth Suri. "Working with AI: Measuring the Applicability of Generative AI to Occupations." arXiv:2507.07935, 2025. https://arxiv.org/abs/2507.07935. Licensed under CC BY 4.0; figures processed for this site.
Anthropic CC BY (the dataset's data; its code is MIT) attribution required

Economic Index 2026-06-26

How people in each occupation used Claude: handed off or worked on together, what for, and each task's share of conversations.

Files: aei_claude_ai_2026-06-26.csv, job_exposure.csv, task_penetration.csv

Credit, as the licence asks: Maxim Massenkoff, Eva Lyubich, Szymon Sacher, Zoe Hitzig, Shaoyi Zhang, Ryan Heller, and Peter McCrory. "Anthropic Economic Index report: Cadences." Anthropic, June 26, 2026. https://www.anthropic.com/research/economic-index-june-2026-report. Data licensed under CC BY; figures processed for this site.
National Center for Education Statistics U.S. Government work (public domain)

CIP 2020 → SOC 2018

Which majors lead to which occupations. It reflects expert judgment, not counts of graduates.

Files: CIP2020_SOC2018_Crosswalk.xlsx
IPUMS CPS, University of Minnesota IPUMS terms of use: cite; do not redistribute the microdata attribution required

IPUMS CPS 13.0, ASEC 2003-2025 (extract cps_00001)

Moves between broad fields from one year to the next. Only our summary table is shared, never the survey records.

Files: cps

Credit, as the licence asks: Sarah Flood, Miriam King, Renae Rodgers, Steven Ruggles, J. Robert Warren, Daniel Backman, Etienne Breton, Grace Cooper, Julia A. Rivera Drew, Stephanie Richards, David Van Riper, and Kari C.W. Williams. IPUMS CPS: Version 13.0 [dataset]. Minneapolis, MN: IPUMS, 2025. https://doi.org/10.18128/D030.V13.0
U.S. Census Bureau U.S. Government work (public domain)

NAICS 2022

Matching industry words in job titles (“venture capital”) to industries.

Files: 2022_NAICS_Index_File.xlsx
U.S. Bureau of Labor Statistics U.S. Government work (public domain)

OEWS May 2025, national industry-specific

Industries named in job titles, and which industry employs most of each occupation.

Files: oesm25in4.zip
U.S. Bureau of Labor Statistics U.S. Government work (public domain)

OEWS May 2025, state estimates

Pay and jobs for each occupation in each state.

Files: oesm25st.zip
Schubert, Stansbury and Taska (Burning Glass Technologies resume data) Used with the authors' permission, on condition of citing the working paper attribution required

BGT occupational mobility data, v. 2021/01/05

Where people in each occupation moved next.

Files: research

Credit, as the licence asks: Gregor Schubert, Anna Stansbury, and Bledi Taska. "Employer Concentration and Outside Options." Working paper, SSRN 3599454. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3599454
U.S. Bureau of Labor Statistics U.S. Government work (public domain)

SOC 2010 to 2018 crosswalk (November 2017)

Putting the resume data's 2010 occupation codes on today's.

Files: soc_2010_to_2018_crosswalk.xlsx
Eloundou, Manning, Mishkin and Rock (OpenAI) MIT (Copyright (c) 2024 OpenAI) attribution required

GPTs-are-GPTs main (Science, 2024)

Which of each occupation's tasks AI could halve the time on (human ratings from 2023).

Files: full_labelset.tsv

Credit, as the licence asks: Tyna Eloundou, Sam Manning, Pamela Mishkin, and Daniel Rock. "GPTs are GPTs: Labor market impact potential of LLMs." Science 384, no. 6702 (2024): 1306-1308. https://doi.org/10.1126/science.adj0998. Task ratings from github.com/openai/GPTs-are-GPTs, Copyright (c) 2024 OpenAI, used under the MIT License.
U.S. Census Bureau U.S. Government work (public domain)

American Community Survey 2024 1-year Public Use Microdata Sample

The education people in each occupation hold, its unemployment rate, and full-time pay by highest degree. Only our summary table is shared, never the survey records.

Files: csv_pus.zip, 2018-occupation-code-list-and-crosswalk.xlsx
U.S. Bureau of Labor Statistics U.S. Government work (public domain)

Job Openings and Labor Turnover Survey (the latest 12 months in the download)

Hires per job opening, over the past year, in the industry that employs most of each occupation.

Files: jt.series, jt.data.2.JobOpenings, jt.data.3.Hires
U.S. Bureau of Labor Statistics U.S. Government work (public domain)

Occupational Requirements Survey, reference year 2023 complete dataset (final)

The share of jobs in each occupation that require a license, a certification or an educational certificate.

Files: 2023-complete-dataset.xlsx
U.S. Department of Labor, Office of Apprenticeship U.S. Government work (public domain)

Approved apprenticeable occupations (undated list; the manifest records the file saved)

Whether the Department of Labor lists registered apprenticeships in an occupation.

Files: approved-occupations.xlsx

O*NET is a trademark of the U.S. Department of Labor, Employment and Training Administration. PathMosaic is not endorsed by any of the organisations listed.