CoChat
Sign Up
Cited by 107Labor market dynamics and wage inequalityDigital Economy and Work TransformationEthics and Social Impacts of AI

The unequal adoption of ChatGPT exacerbates existing inequalities among workers

Open access
Humlum, Anders, Vestergaard, Emilie·2025·Proceedings of the National Academy of Sciences
DOI 10.1073/pnas.2414972121PMC PMC11725873
Open this paper in CoChat — free account

Abstract

We study the adoption of ChatGPT, the icon of Generative AI, using a large-scale survey linked to comprehensive register data in Denmark. Surveying 18,000 workers from 11 exposed occupations, we document that ChatGPT is widespread, especially among younger and less-experienced workers. However, substantial inequalities have emerged. Women are 16 percentage points less likely to have used the tool for work. Furthermore, despite its potential to lift workers with less expertise, users of ChatGPT earned slightly more already before its arrival, even given their lower tenure. Workers see a substantial productivity potential in ChatGPT but are often hindered by employer restrictions and a perceived need for training.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1073/pnas.2414972121

Cites

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • We used a single LLM (ChatGPT-4o) as the basis for AI-assisted workflows.

    “Fourth, we used a single LLM (ChatGPT-4o) as the basis for AI-assisted workflows, considering its popularity among general and non-expert users [34, 39].”

    From How University Disability Services Professionals Write Image Descriptions for HCI Figures Using Generative AI (2026) · cites this paper as [34]

    Materials and Methods describes ChatGPT as the leading generative AI chatbot, with 80–90% of chatbot website traffic. That supports the popularity rationale, but the paper does not establish popularity specifically for GPT-4o or among non-expert users. The citing authors’ choice to use one model is their own method, not a finding attributed to this source. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • Denmark's workers are among the earliest adopters of Generative AI, with usage rates similar to the U.S.

    “Danish workers have been at the forefront of Generative AI adoption, with take-up rates comparable to those in the United States”

    From Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI (2025) · cites this paper as Humlum and Vestergaard, 2025

    The Conclusion reports rapid take-up of ChatGPT among Danish workers in the occupations studied, and the Abstract says 41% had used it for work about a year after launch. Our copy does not establish that Danish workers were among the earliest adopters or compare their take-up rates with those in the United States. (LLM, conf=0.83)

    Experimental — accuracy not yet measured

  • No other new technology has been adopted by workers as quickly as AI chatbots.

    “AI chatbots have seen the fastest worker take-up of any new technology”

    From Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI (2025) · cites this paper as Humlum and Vestergaard, 2025

    The Abstract (page unavailable) reports that 41% of surveyed workers in exposed occupations had used ChatGPT for work about a year after launch, describing its adoption as a landmark event. The Conclusion (page unavailable) also calls the take-up rapid. Neither passage establishes that AI chatbots had the fastest worker adoption of any new technology. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • In the 2023 survey round, women were significantly more likely than men to cite lack of training as a barrier to ChatGPT adoption.

    “The effectiveness of training aligns with findings from the 2023 survey round, in which women were significantly more likely to cite lack of training as a barrier to ChatGPT adoption (Humlum and Vestergaard, 2025).”

    From Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI (2025) · cites this paper as Humlum and Vestergaard, 2025

    The Gender Gaps section says women were more likely to report training-related adoption barriers and that 45% of the affected women said they needed training. It does not explicitly establish statistical significance for that comparison. The Survey section dates data collection to November 2023–January 2024, rather than a distinct 2023 survey round. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • The 2023 survey used a similar protocol with minor modifications.

    “The 2023 survey followed a similar protocol, with minor modifications; see Humlum and Vestergaard (2025) for details.”

    From Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI (2025) · cites this paper as Humlum and Vestergaard (2025)

    The Materials and Methods > Survey section describes the survey conducted from November 2023 to January 2024 and points to details of its questionnaire and survey blocks. It does not compare that protocol with another survey or describe the claimed minor modifications. (LLM, conf=0.78)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

No citation lists for this paper

No citation lists for this paper yet — checked again daily.

References 14 · Cited by 107

Research works printed but not linked: 5

  1. [1]T. Eloundou, S. Manning, P. Mishkin, D. Rock, GPTs are GPTs: Labor market impact potential of LLMs. Science 384, 1306– 1308 (2024). 38900883
  2. [3]E. Brynjolfsson, D. Li, L. Raymond, Generative AI at work. arXiv [Preprint] (2024). https://arxiv.org/abs/2304.11771 (Accessed 13 December 2024) · arXiv
  3. [8]N. Otis, R. P. Clarke, S. Delecourt, D. Holtz, R. Koning, The uneven impact of Generative AI on entrepreneurial performance. SSRN (2024). https://ssrn.com/abstract=4671369. Accessed 13 December 2024
  4. [11]A. L. Eisfeldt, G. Schubert, M. B. Zhang, B. Taska, The labor impact of Generative AI on firm values. SSRN (2024). https://ssrn.com/abstract=4436627. Accessed 13 December 2024.
  5. [13]C. R. Sunstein, R. H. Thaler, Nudge: Improving Decisions about Health, Wealth, and Happiness (Yale University Press, 2008).

Authors
Humlum, A., & Vestergaard, E.
Affiliations
  • University of Chicago (US)
  • University of Copenhagen (DK)
Corresponding author
  • Humlum, Anders
  • Vestergaard, Emilie
ORCID
Vestergaard, Emilie
Venue
Proceedings of the National Academy of Sciences
Publisher
National Academy of Sciences
Published
2024-12-30
Volume
122
Issue
1
Pages
e2414972121-e2414972121
ISSN
  • 0027-8424 (ISSN-L)
  • 1091-6490
DOI
10.1073/pnas.2414972121
PMC
PMC11725873
OpenAlex
W4405896613
Cited by
107
References
14
Citations by year
2026: 62 · 2025: 44 · 2024: 1
FWCI
37.93
Type
journal article
Language
en
Open access
Licence
Locations
  • Proceedings of the National Academy of Sciences (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Research at the University of Copenhagen (University of Copenhagen) (publishedVersion, cc-by)
  • Knowledge@UChicago (University of Chicago) (submittedVersion)
  • University of Chicago (cc-by)
Funders
  • University of Chicago
  • Booth School of Business, University of Chicago
Topics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Labor market dynamics and wage inequality
  • Social Sciences › Sociology and Political Science › Digital Economy and Work Transformation
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
Keywords
ChatGPT, generative AI adoption, labor market inequality, employee experience, AI training needs, productivity, occupational exposure, Denmark
MeSH terms
Adult, Denmark, Employment, Female, Humans, Male, Middle Aged, Occupations, Surveys and Questionnaires, Socioeconomic Factors

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review