The Legitimation-Validation Paradox in AI-Driven Talent Acquisition : A Critical Systematic Review

AI-based recruitment tools are increasingly used to screen, match, assess, and select candidates, yet the evidence supporting their predictive validity remains less developed than their organizational legitimacy and scholarly visibility. This article examines this tension as a legitimation-validation paradox, defined as a condition in which technologies become institutionally accepted while their core performance claims remain insufficiently substantiated. The study conducts a critical systematic review with a theoretical aim, following the PRISMA 2020 protocol and applying controversy mapping rather than conventional thematic synthesis. The corpus comprises 110 publications, including studies identified within the 2015–2024 search window and foundational works retrieved through backward citation searching. Searches were conducted in Scopus, Web of Science, EBSCO Business Source, PsycINFO, IEEE Xplore, JSTOR, Google Scholar, ProQuest Dissertations, and Cairn.info. Grey literature and trade publications were excluded from the database-search evidence base, except when used as contextual or foundational sources identified through citation tracking. The analysis reveals a persistent asymmetry between relatively well-documented efficiency claims and weaker, contested, or insufficiently replicated evidence concerning criterion validity, fairness, and predictive performance. By distinguishing operational efficiency from predictive validity and fairness, the article reorients the AI recruitment debate toward the prior question of independent validation. The findings should be interpreted cautiously, given the predominantly Western and Anglophone composition of the reviewed corpus.