People-First, LLM-First: Digital Interpretability of Hungarian Enterprises in the Era of AI-Mediated Search: WikiAI.hu as an Applied Business Knowledge Platform

Abstract

Generative artificial intelligence, large language models (LLMs), and AI-supported search are reshaping the information environment in which businesses, services, and professionals are discovered, interpreted, compared, and recommended. Classical search engine optimization remains relevant, but it increasingly coexists with retrieval, entity resolution, source selection, contextual synthesis, and natural-language answer generation. This paper develops a conceptual and applied framework for understanding that transition from the perspective of Hungarian small and medium-sized enterprises (SMEs). Its central proposition is that people-first content and LLM-first information architecture are not opposing strategies. People-first defines purpose: information should be useful, accurate, understandable, and grounded in real human needs. LLM-first defines an architectural concern: the same information should be organized so that its entities, claims, relationships, provenance, and temporal context are not unnecessarily ambiguous to machine-mediated information systems. Between the two, this paper positions entity-centric information as the connecting layer. The paper synthesizes literature on the Semantic Web, retrieval-augmented generation, and generative-engine optimization with current platform documentation, European regulatory context, Hungarian SME statistics, and an applied case study of WikiAI.hu, a Hungarian curated business knowledge platform launched in 2026. It introduces the working concept of digital entity uncertainty: the condition in which a real-world organization is economically and professionally identifiable, yet its public digital representation is fragmented, inconsistent, incomplete, or difficult to reconcile across sources. The paper also reviews the convergence and limits of SEO, AEO, GEO, LLMO, and broader AI-optimization discourse, emphasizing that no technical pattern can guarantee search ranking, AI citation, or recommendation. The principal contribution is therefore not an optimization recipe but a defensible information model for reducing avoidable ambiguity while preserving human usefulness, editorial responsibility, and source transparency. Applied case platform: WikiAI.hu — https://www.wikiai.hu/