Talking Technology, Not Customers: Vocabulary Decoupling in Corporate Artificial Intelligence Disclosure

Artificial intelligence (AI) is widely described as reshaping marketing administration through personalization, predictive analytics, and automated customer engagement. Whether firms describe AI as a customer-facing marketing capability or as a technology and a risk to be disclosed remains an open empirical question. This study uses mandated disclosure as an observational window on that question. Drawing on a balanced panel of 1100 firm-years from 100 U.S.-listed consumer-facing firms in 15 sectors (fiscal years 2015 to 2025), I measure technology vocabulary and marketing vocabulary in Form 10-K filings using phrase-presence indicators from the EDGAR full-text index and graded, section-located counts from the full text of each filing. Three findings emerge. First, AI disclosure diffused rapidly and converged: the share of firm-years disclosing rose from 28.7% in fiscal 2020 to 2022 to 91.7% in 2023 to 2025, and the intensity of AI language rose roughly twentyfold between 2021 and 2025. Second, within firms and net of common year shocks, adopting AI vocabulary was not accompanied by a substantively meaningful increase in marketing vocabulary (standardized β = 0.02, 95% CI [−0.09, 0.13]); graded and staggered-adoption estimates leave room for at most a small, lagged association. Only 3.2% of AI sentences written in 2023 to 2025 used marketing vocabulary, and AI language migrated from the Business section toward Risk Factors and MD&A. Third, first disclosure was ordered by sector, with digital-commerce and financial firms preceding hospitality and travel firms. I interpret the pattern as vocabulary decoupling: AI entered the technology-and-risk register of the annual report largely without being connected to the register in which firms account for value delivered to customers. The findings concern disclosure, not deployed capability.