CoChat

Generative Artificial Intelligence in Healthcare: Automation Bias, Deskilling, and Cognitive Implications – A Systematic Review

Open access
Al-Anezi, Fahad M · 2026 · Journal of Healthcare Leadership

The DOI matches its Crossref record; 51 of the 51 references it cites were found in a registry or an index; 27 of the 98 sources its claims cite were checked for support.

Abstract

Background: Generative artificial intelligence (GenAI) is rapidly transforming eHealthcare, promising substantial gains in diagnostic accuracy, workflow efficiency, and decision support, yet raising concerns about automation bias and clinician deskilling that may erode core diagnostic expertise and professional judgment over time. Aim and Objectives: This systematic review investigates the dual impact of GenAI in eHealthcare, focusing on how it enhances clinical efficiency and decision support while potentially diminishing clinicians' diagnostic expertise and professional judgment. Methods: A systematic search was conducted across PubMed, Scopus, Web of Science, and Google Scholar. A total of 11,269 records were identified, and 29 studies met inclusion criteria following PRISMA 2020 guidelines. Studies were synthesized using a narrative approach due to heterogeneity in design, clinical domains, and AI system maturity. Results: Across the included studies, recurrent themes included automation bias (reported in 10 studies), concerns regarding clinician deskilling (9 studies), and impacts on diagnostic reasoning (9 studies). Evidence was predominantly observational, experimental simulation-based, or conceptual in nature. Findings reveal that GenAI significantly improves diagnostic accuracy, workflow efficiency, and decision quality across radiology, telehealth, and education domains. However, overreliance introduces risks of automation bias, cognitive deskilling, and loss of interpretive autonomy. The review identifies key mitigation strategies, including human-in-the-loop frameworks, explainable AI (XAI), ethical governance, and continuous clinician reskilling. Theoretically, the results redefine the clinician-AI relationship as a dynamic cognitive partnership, while practically they emphasize responsible integration through education and regulation. Conclusion: Sustainable adoption of GenAI demands balanced implementation-leveraging its analytical capabilities without compromising human judgment, clinical reasoning, or professional accountability.

Overview

CoChat Summary

TL;DR

This systematic review synthesizes 29 studies on generative AI in eHealthcare, finding measurable gains in diagnostic accuracy, workflow efficiency, and decision support alongside risks of automation bias, deskilling, and reduced interpretive autonomy. It argues for human-in-the-loop, explainable, and ethically governed integration to preserve clinician judgment.

Key findings

  • Across 29 included studies, recurrent themes were automation bias (10 studies), clinician deskilling concerns (9 studies), and impacts on diagnostic reasoning (9 studies).
  • GenAI improved diagnostic/prognostic efficiency by 40–45% in some analyses; ChatGPT-4 was 87% accurate in clinical education/reasoning versus 48% for 3.5, and Beheshti et al. reported 73% accuracy across 128 evaluations.
  • Tool-augmented models reduced medication-dose mistakes by as much as 88%, and one study documented a 30% decrease in diagnostic latency for endocrine disorders.
  • Risks included diminished interpretive engagement among endoscopists, with reductions up to 20% in unassisted colonoscopy diagnosis accuracy after AI adoption; GPT-4 failed to achieve the accurate final diagnosis in 16% of instances in one study.

Methods

  • Systematic review following PRISMA 2020, searching PubMed, Scopus, Web of Science, ScienceDirect, IEEE Xplore, and Google Scholar for January 2015–June 2025; 11,269 records were identified and 29 studies met inclusion criteria.
  • Included studies were empirical, conceptual, systematic, or mixed-method, spanning radiology, neurology, endoscopy, telehealth, surgery, and medical education; sample sizes ranged from small experimental cohorts to large syntheses.
  • Due to heterogeneity in design, clinical domains, AI maturity, and outcomes, quantitative meta-analysis was not appropriate; a structured narrative synthesis was used, with CASP quality appraisal (14 high quality ≥8/10; 15 moderate 6–7/10; none ≤5).

Limitations

  • The paper states that excluding non-English publications (n = 103) may introduce language bias and miss relevant global contributions.
  • The paper states that evidence remains predominantly short-term, heterogeneous, and observational, with limited direct longitudinal evidence of sustained professional deskilling and a need for longitudinal, real-world, threshold-based studies.
  • Evident from the text: common included-study limitations were simulation-based designs, absence of long-term outcome assessment, and limited real-world implementation data; the review also relied on narrative synthesis rather than meta-analysis.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.2147/jhl.s590498

Cites

0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Screening and selection process using PRISMA.
Figure 1 Screening and selection process using PRISMA.CC BY-NC · PubMed Central

Verification

Claims the paper makes

14 Valid · 1 Not valid · 51 Not checked · 38 Own results and interpretations

40 more claims · 5 with their cited sources checked

Create a free account to see all 104 claims and to verify the 2 references still open.

See all claims — free account
  • It may state two separate claims as one. 33
  • Probably not a substantial claim. 30
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Self-check

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

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

The publisher hasn’t deposited this paper’s reference list.

Cited by

See all 6

Metadata

Authors
Al-Anezi, F. M.
Venue
Journal of Healthcare Leadership
Published
2026-07
Volume
Volume 18
Pages
1-13
ISSN
1179-3201 (ISSN-L)
OpenAlex
W7167801398
Cited by
6
References
0
Citations by year
2026: 6
FWCI
3.66
Type
journal article
Language
en
Open access
Locations
  • Journal of Healthcare Leadership (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by-nc)
Topics
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Health Sciences › Medicine › Family Practice › Clinical Reasoning and Diagnostic Skills
  • Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
Keywords
generative AI, diagnostic reasoning, clinical decision support, overreliance on AI, human-in-the-loop, explainable AI, AI governance, radiology, telemedicine, medical education

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Generative Artificial Intelligence in Healthcare: Automation Bias, Deskilling, and Cognitive Implications – A Systematic Review. https://app.cochat.ai/papers/10.2147/jhl.s590498
The paper
Al-Anezi, F. M. (2026). Generative Artificial Intelligence in Healthcare: Automation Bias, Deskilling, and Cognitive Implications – A Systematic Review. Journal of Healthcare Leadership. https://doi.org/10.2147/jhl.s590498

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.