AI has reshaped labor markets unevenly: while productivity and output gains are common, benefits are highly concentrated—fueling factor‑income polarization, rising skill demands, and elevated displacement risks, with much of the evidence coming from a few lead economies.

Main Finding

A systematic review of 180 representative articles (published in leading journals, 2000–Mar 2026) finds that AI has produced a systemic, heterogeneous shock to labor markets. Research is clustered around major technological breakthroughs—most recently generative AI—and concentrated geographically (notably the U.S. and China). The literature has moved from static, aggregate exposure measures toward dynamic, multidimensional frameworks that combine task content with micro-level worker characteristics. AI’s net effects are mixed and asymmetric: aggregate productivity gains coexist with factor‑income polarization, rising skill demands coexist with displacement risks, and occupational insecurity and skill mismatches are widespread.

Key Points

• Research drivers: Publication patterns and topics track major AI advances (especially generative AI), producing research surges after key breakthroughs.
• Geographic concentration: Empirical and theoretical work is disproportionately produced in a few economies (e.g., United States, China), limiting cross-country coverage and comparability.
• Measurement evolution: The field has shifted from coarse, static measures of "automation exposure" toward dynamic, multidimensional vulnerability frameworks that integrate task content, worker skills, and job/firm heterogeneity.
• Mechanisms: Labor‑market impacts emerge from interactions among technological attributes (capabilities, complementarities/substitutability), individual endowments (skills, education), organizational heterogeneity (firm size, adoption choices), and institutional context (labor regulations, social insurance).
• Distributional outcomes: Productivity and output gains are common, but benefits are uneven—evidence of factor‑income polarization (winners and losers), increased displacement risks for some occupations, skill mismatches, and greater occupational insecurity.
• Research gaps: Persistent data limitations, identification challenges, and limited cross‑country comparability constrain causal inference and policy guidance.

Data & Methods

• Corpus: 180 representative articles from leading journals, covering 2000 through March 2026.
• Approaches used in the review:
  • Bibliometric analysis to map publication patterns, topic emergence, geographic distribution, and links to technological milestones.
  • Qualitative synthesis to extract transmission mechanisms, empirical findings, measurement approaches, and policy discussions.
• Empirical strategies reported in the literature (synthesized): task‑based exposure indices, matched employer–employee microdata, firm‑level adoption studies, difference‑in‑differences and instrumental variables where available, structural models for general equilibrium and distributional effects.
• Trend identification: Documented methodological progression from aggregate exposure metrics to richer, task‑and-worker‑level vulnerability frameworks and more dynamic, context‑sensitive analyses.

Implications for AI Economics

• For research:
  • Prioritize richer microdata (matched worker–job–firm panels), international harmonization, and longitudinal designs to trace dynamics of displacement, reallocation, and skill accumulation.
  • Develop identification strategies that exploit plausibly exogenous variation in AI capabilities/adoption and combine structural and reduced‑form approaches for counterfactuals.
  • Expand geographic coverage beyond high‑income leaders to understand heterogenous institutional responses and global spillovers.
• For policy:
  • Design dynamic, context‑sensitive policies: combine active labor‑market programs (retraining, mobility support), stronger social insurance against transition risk, and measures to encourage complementary skill formation.
  • Anticipate distributional tradeoffs: leverage taxation, redistribution, and institutionally tailored interventions to address factor‑income polarization and uneven gains across workers and firms.
  • Support firm-level adoption strategies that promote worker complementarities (job redesign, on‑the‑job training) to capture productivity gains while mitigating displacement.
• For practitioners and institutions:
  • Monitor technological capability trends (e.g., generative models) and their sectoral footprints to target reskilling and regulatory attention.
  • Promote data sharing and cross‑country benchmarking initiatives to improve policy learning and evidence synthesis.