Communicating Artificial Intelligence: A Longitudinal Study of AI Disclosure in the Annual Reports of Large EU-Listed Firms

Sammanfattning

This study examines how artificial intelligence (AI) is positioned as a strategic priority in the annual reports of large EU-listed firms, and how this positioning has shifted alongside the development of the EU regulatory framework for AI. Drawing on institutional theory, legitimacy theory, and the resource-based view, the study analyses 220 firm-year observations across 20 firms from 2015 to 2025 using longitudinal keyword-based content analysis of approximately 22 million words. The findings reveal an 82-fold increase in cross-firm average AI disclosure intensity over the study period, structured by three distinct phases in which the 2022–2023 transition represents the sharpest turning point, driven primarily by the public emergence of generative AI. Firm-level variation is persistent: technology-native leaders average 1.520 AI mentions per 1,000 words in 2025 while persistent low disclosers average 0.110, and the gap between groups widened rather than narrowed. Act-specific governance vocabulary, including Responsible AI, Human oversight, and AI governance, emerges exclusively after 2022 with timing corresponding to the EU AI Act's legislative milestones, constituting indirect evidence of regulatory influence on disclosure composition. However, ethics and governance language has not kept pace with the surge in AI language post-2022: the Ethics to AI ratio falls from 6.513 in 2022 to 1.319 in 2025, indicating that the dominant mode of AI positioning remains strategic and competitive rather than governance-oriented. The study offers two contributions: a replicable longitudinal methodology for tracking disclosure shifts over time, and evidence that mimetic and coercive isomorphic pressures operate through mechanisms that can be separated empirically in vocabulary composition, not just volume.