When AI hurts the pros: expertise-fixation costs and platform deployment of generative AI in subscription markets

Abstract

Purpose This study examines when subscription-based content platforms should deploy generative AI, given the heterogeneous AI-induced Einstellung effects on novice and expert creators. It identifies conditions under which deployment improves or undermines platform performance. Design/methodology/approach A calibrated two-sided market model characterizes equilibrium outcomes under AI and non-AI regimes, derives closed-form deployment thresholds, and tests robustness to heterogeneous expert adoption and positive novice baseline quality. Findings AI deployment is governed by a cost threshold, with welfare effects depending on implementation costs and standalone utility for a consumer joining the platform. Low-cost deployment can benefit platforms, creators, and consumers, whereas high costs may reduce profitability and stakeholder welfare by amplifying expert-side cognitive friction. Maximizing AI adoption is therefore not always optimal, even when AI raises average productivity. Research limitations/implications The analysis treats creator quality as exogenous and considers only one platform. The core deployment logic nevertheless remains robust to heterogeneous expert adoption and positive novice baseline quality. Practical implications Before deployment, managers should assess the composition of creators, implementation costs, and consumer demand. Enhancing AI-workflow compatibility and using targeted strategies can reduce expert-side friction and expand the conditions under which AI creates ecosystem value. Social implications Indiscriminate AI deployment may discourage expert participation and reduce content quality and consumer welfare, despite increasing the overall content supply. Originality/value This study offers a cognition-aware deployment framework by incorporating heterogeneous creator responses into a two-sided market model. It identifies thresholds and welfare regimes showing when AI creates mutual gains and when strategic restraint is preferable.