Academic economics on AI is clustered around supply chains, labor markets and large language models, with a sharp rise in LLM studies after ChatGPT; crucial policy questions—measuring AI-driven growth, evaluating industrial policy, and assessing regulatory and cross-border effects—remain underexplored.

Mapping the Landscape of the Economics of AI Literature: Gaps and Opportunities for Research and Policy

Lucy Hampton, J. L. Poquiz · Fetched March 15, 2026 · Journal of economic surveys (Print)

semantic_scholar review_meta n/a evidence 7/10 relevance DOI Source

A topic-modeling review of over 4,600 papers finds AI-economics research concentrated on supply chains, labor markets, and LLMs (with an LLM surge after ChatGPT), while policy-relevant areas—like measuring AI-driven growth, industrial-policy impacts, regulatory effects, and cross-border governance—are comparatively neglected.

This paper investigates the evolving landscape of research on the economics of artificial intelligence (AI). Using topic modeling on a corpus of over 4,600 academic papers, we identify the dominant themes that have shaped the literature to date and highlight areas that remain relatively underexplored. In addition, we assess the extent to which current research trajectories align with policy priorities. Our findings show significant research concentration on AI applications in supply chains, labor markets, and large language models (LLMs), especially following ChatGPT's release. However, comparing these topics to national AI strategies and legislation across regions, we identify notable gaps, particularly in measuring AI‐driven economic growth, effective industrial policy, and the economic impacts of risk‐based AI regulations. Global AI governance, regulatory fragmentation, and the effects of privacy laws on market competition in this context are also under‐studied. By mapping these trends and gaps, the study offers guidance for future research and for policymakers navigating AI's economic and regulatory landscape.

Summary

Main Finding

Research on the economics of AI has concentrated on a few themes—particularly AI in supply chains, labor markets, and large language models (LLMs)—with a sharp uptick in attention to LLMs after ChatGPT’s release. When compared to national AI strategies and legislation across regions, important policy-relevant areas remain underexplored, notably measurement of AI-driven economic growth, the effectiveness of industrial policy, economic effects of risk-based AI regulation, global governance, regulatory fragmentation, and how privacy laws affect competition.

Key Points

Corpus and approach: Topic modeling applied to a corpus of over 4,600 academic papers to identify dominant research themes and temporal shifts.
Dominant themes: High concentration on supply chain applications, labor-market effects, and LLMs; LLM-related research accelerated notably after ChatGPT.
Policy misalignment: Many topics prioritized in national AI strategies and legislation receive relatively little attention in the academic economics literature.
Notable gaps:
- Measurement and attribution of AI-driven economic growth.
- Empirical assessment of industrial policy aimed at AI-related industries.
- Economic impacts of risk-based regulatory approaches to AI.
- Research on global AI governance and the consequences of regulatory fragmentation.
- Effects of privacy and data-protection laws on market competition in AI contexts.
Value of mapping: The topic-map highlights where academic efforts cluster and where additional evidence would most help policymakers.

Data & Methods

Data: A curated corpus of more than 4,600 academic papers on AI economics.
Methodology: Unsupervised topic modeling to identify dominant themes and their evolution over time; comparative analysis between identified research topics and the priorities expressed in national AI strategies and legislation across regions.
Temporal analysis: Detection of shifts in topic prevalence, including post-ChatGPT increases in LLM-related studies.
Limitations (implicit from method): Topic models reveal thematic concentration but do not by themselves establish causal findings; coverage depends on corpus selection and the policy document sample used for alignment checks.

Implications for AI Economics

Research agenda:
- Prioritize empirical work measuring AI’s contribution to aggregate productivity and growth, including better identification strategies and new measurement tools.
- Evaluate industrial policy interventions targeted at AI (subsidies, R&D support, procurement) using firm- and region-level data and causal inference methods.
- Study the economic consequences of different regulatory approaches (e.g., risk-based regulation), including compliance costs, innovation incentives, and distributional effects.
- Investigate cross-border issues: how regulatory divergence affects multinational firms, data flows, and competitive dynamics.
- Analyze interactions between privacy/data-protection rules and market concentration or competition in AI-driven markets.
For policymakers:
- Commission and support research that fills the identified empirical gaps to inform policy design and ex ante impact assessment.
- Foster data-sharing initiatives and access to firm-level and administrative data to enable robust evaluation.
- Coordinate internationally to reduce harmful fragmentation and to better understand cross-border spillovers of AI policy.
Methodological recommendations:
- Combine topic-mapping with targeted systematic reviews and meta-analyses in underexplored policy areas.
- Use quasi-experimental and structural approaches to isolate AI effects on growth, employment, and market structure.
- Encourage interdisciplinary collaborations (economics, law, political science, computer science) to address complex regulatory and governance questions.

Assessment

Paper Typereview_meta Evidence Strengthn/a — This is a descriptive, corpus-level topic-mapping and comparative analysis of the literature and policy documents rather than an empirical paper that tests causal hypotheses or estimates effects. Methods Rigormedium — The study uses a large, curated corpus (>4,600 papers) and standard unsupervised topic-modeling and temporal analysis methods, which are appropriate for mapping research trends; however, results depend on corpus selection, model specification and labeling choices, and there is limited validation of topic assignments or of the policy-document sampling, and no causal identification. SampleA curated corpus of more than 4,600 academic papers on the economics of AI (including topics like supply chains, labor markets, and LLMs) analyzed with unsupervised topic modeling and temporal trend analysis; complemented by a sample of national AI strategies and legislation across regions for comparative alignment checks (temporal coverage includes the period after the release of ChatGPT). Themeslabor_markets governance GeneralizabilityFindings depend on the curated corpus composition (language, publication type, inclusion/exclusion criteria) and may omit gray literature or recent preprints., Topic-modeling outputs are sensitive to model choice, number of topics, and labeling, which can affect thematic classification., Temporal inferences (e.g., post-ChatGPT surge) may be influenced by publication lags and preprint growth, not just research interest., Policy-document comparison may be limited by the selection of national strategies/legislation and regional representation., The mapping describes research concentrations and gaps but does not generalize to causal statements about AI’s economic impacts.

Claims (8)

Claim	Direction	Confidence	Outcome	Details
The study uses topic modeling on a corpus of over 4,600 academic papers to identify the dominant themes in the economics of AI literature. Research Productivity	null_result	high	identified topics / dominant themes (topic prevalence across the corpus)	n=4600 0.04
There is significant research concentration on AI applications in supply chains, labor markets, and large language models (LLMs). Research Productivity	positive	medium	prevalence (share or weight) of papers categorized under supply chains, labor markets, and LLM-related topics	n=4600 0.02
Research on large language models (LLMs) has increased especially after the release of ChatGPT. Research Productivity	positive	medium	time-series change in share of papers on LLMs (increase in LLM-topic prevalence post-ChatGPT)	n=4600 0.02
There are notable gaps in the literature in measuring AI-driven economic growth. Research Productivity	negative	medium	coverage/prevalence of studies measuring AI-driven economic growth	n=4600 0.02
Research on effective industrial policy for AI is relatively underexplored. Research Productivity	negative	medium	coverage/prevalence of research on AI-related industrial policy	n=4600 0.02
The economic impacts of risk-based AI regulations are under-studied in the current literature. Research Productivity	negative	medium	coverage/prevalence of studies examining economic impacts of risk-based AI regulations	n=4600 0.02
Global AI governance, regulatory fragmentation, and the effects of privacy laws on market competition are under-studied areas. Research Productivity	negative	medium	coverage/prevalence of research on global AI governance, regulatory fragmentation, and privacy laws' effects on competition	n=4600 0.02
By mapping trends and gaps in the literature, the study offers guidance for future research and for policymakers navigating AI's economic and regulatory landscape. Governance And Regulation	positive	low	qualitative guidance (recommendations) for future research and policy priorities	n=4600 0.01