AI-powered travel services increasingly mediate destination discovery, itinerary formation, cultural interpretation, and interactions with local providers. Yet many platforms optimise individual transactions while leaving recommendations opaque, heritage representations epistemically ambiguous, accessibility uneven, and local participation vulnerable to ranking inequalities. This study develops Kolehposhti as a conceptual socio-technical architecture for responsible smart tourism. Using a design-science orientation, it combines an integrative, design-oriented synthesis of 44 scholarly and authoritative sources with a structured analysis of the project brief. The analysis applies domain coding, requirement derivation, and requirements-to-module traceability; the project brief is treated as a design input rather than empirical evidence. Six interdependent requirement domains emerge: context-sensitive decision support, explainable human-AI interaction, evidence-labelled immersive heritage, multilingual and accessible mediation, equitable local-provider participation, and embedded governance. These requirements are integrated into six modules: a traveller-profile and consent layer, an itinerary and recommendation engine, a multilingual interface, an immersive heritage layer, a verified local-provider ecosystem, and a governance and accountability layer. The framework extends smart-tourism research by operationalising explainability, cultural provenance, contestability, and exposure fairness as architectural requirements rather than optional features. It does not claim demonstrated usability or impact. Instead, it provides a traceable design proposition and a testable evaluation agenda for future prototype studies with travellers, disabled users, local providers, guides, heritage specialists, and destination stakeholders.

smart tourism; responsible artificial intelligence; explainable recommendation; cultural heritage; local participation