This chapter examines Tagore.ai, a generative AI system developed by Factly Media, as a case of informative AI that foregrounds reliability in conditions of disinformation, information overload, and large language model hallucinations. Situated within the two-tier framework of informative ICTs, the chapter analyses how reliability is operationalized by design through curated public-interest datasets, constrained retrieval, provenance disclosure, and institutional stewardship. Using interviews, document analysis, observation of the live system, and structured querying, the chapter shows how Tagore transforms dispersed official records into verifiable and accessible civic knowledge. It also discusses social entropy as an analytical lens to explain how informativeness can emerge from archival and institutional plurality rather than participatory input. The chapter argues that reliability-by-design is a necessary condition for information integrity in politically contested, multilingual, and AI-mediated information environments.