This article surveys the emerging empirical literature on the impact of artificial intelligence (AI) – and particularly generative AI – on labor market outcomes. Despite widespread public debate, the evidence remains inconclusive and highly heterogeneous. Drawing on both individual studies and three recent meta- analyses, the paper identifies four broad channels of impact: employment, wages, productivity, and inequality. It highlights contrasts in methodology, geographical coverage, sectoral focus, and definitions of AI, emphasizing the need to distinguish between task-based automation and general-purpose technologies such as large language models. Across the literature, no consistent evidence has yet emerged of systemic labor market disruption attributable to generative AI. While displacement occurs in certain occupations – especially within digital freelance markets – other sectors exhibit productivity gains or wage premia. Recent meta-analytic studies confirm the absence of statistically significant average effects of AI adoption on employment or wages and detect no robust signs of publication bias. These findings reflect the still-emergent diffusion of AI and its uneven adoption across industries and countries. The article concludes by identifying priorities for future research.
Questo articolo esamina la letteratura empirica emergente sugli effetti dell’intelligenza artificiale (IA) – e in particolare dell’IA generativa – sui risultati occupazionali. Nonostante l’ampio dibattito pubblico, le evidenze disponibili restano inconcludenti ed eterogenee. Basandosi su studi individuali e tre recenti meta-analisi, il contributo identifica quattro principali categorie di effetti: occupazione, salari, produttività e disuguaglianze. Vengono messi in luce contrasti nelle metodologie, nei contesti geografici e settoriali, nonché nelle definizioni stesse di IA, sottolineando l’importanza di distinguere tra automazione basata sui compiti e tecnologie generaliste come i modelli linguistici di grandi dimensioni. Complessivamente, non emergono evidenze coerenti di una disgregazione sistemica dei mercati del lavoro imputabile all’IA generativa. Se da un lato si registrano effetti di sostituzione in alcune professioni – soprattutto nei mercati digitali freelance – dall’altro si osservano guadagni di produttività o premi salariali in determinati settori. Le più recenti evidenze meta-analitiche confermano l’assenza di effetti medi statisticamente significativi dell’adozione dell’IA su occupazione e salari, e non rilevano segnali robusti di bias di pubblicazione. Questi risultati riflettono la natura ancora emergente della diffusione dell’IA e la sua adozione disomogenea. L’articolo si chiude delineando le priorità per la ricerca futura.
Carbonara, E., Santarelli, E. (2025). The Labor Market Impact of Generative AI: A Critical Survey of the Empirical Literature. RIVISTA TRIMESTRALE DI SCIENZA DELL'AMMINISTRAZIONE, 70(4), 1-30 [10.32049/RTSA.2025.4.04].
The Labor Market Impact of Generative AI: A Critical Survey of the Empirical Literature
Carbonara, E.
Writing – Original Draft Preparation
;Santarelli, E.Writing – Original Draft Preparation
2025
Abstract
This article surveys the emerging empirical literature on the impact of artificial intelligence (AI) – and particularly generative AI – on labor market outcomes. Despite widespread public debate, the evidence remains inconclusive and highly heterogeneous. Drawing on both individual studies and three recent meta- analyses, the paper identifies four broad channels of impact: employment, wages, productivity, and inequality. It highlights contrasts in methodology, geographical coverage, sectoral focus, and definitions of AI, emphasizing the need to distinguish between task-based automation and general-purpose technologies such as large language models. Across the literature, no consistent evidence has yet emerged of systemic labor market disruption attributable to generative AI. While displacement occurs in certain occupations – especially within digital freelance markets – other sectors exhibit productivity gains or wage premia. Recent meta-analytic studies confirm the absence of statistically significant average effects of AI adoption on employment or wages and detect no robust signs of publication bias. These findings reflect the still-emergent diffusion of AI and its uneven adoption across industries and countries. The article concludes by identifying priorities for future research.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


