Data and Audit Materials for: Seeing Is Not Verifying: An Empirical Audit of Provenance Disclosure, Prompt Mechanics, and Epistemic Framing in Viral Generative-Image Nostalgia

Abstract

This dataset and audit-materials repository supports the study “Seeing Is Not Verifying: An Empirical Audit of Provenance Disclosure, Prompt Mechanics, and Epistemic Framing in Viral Generative-Image Nostalgia.” The study examines publicly accessible generative-image records associated with the viral 1980s-style photo trend in India during September 2026. The research focuses on provenance disclosure, procedural disclosure, epistemic framing, platform-level AI-label observability, and the recoverability of information needed to verify how realistic synthetic historical-looking images were produced. The repository contains the study's analytic and audit materials, including the verified primary-record ledger, GINR (Generative Image Nostalgia Record) identifiers, binary-asset and provenance documentation, perceptual-hash (pHash) outputs, substantive coding records, adjudication and construct-validity audit materials, and the final adjudicated analytic dataset. The corpus-assembly process identified 63 candidates retained in the analytic candidate denominator: 22 verified-primary records, 32 candidates pending primary verification, and nine substantive exclusions. The final binary-eligible static analytic frame comprised 14 GINR records representing 15 authentic static assets. The pHash audit comprised 105 pairwise comparisons. Substantive coding was conducted in two separately executed coding passes against the frozen coding framework, followed by adjudication. The repository preserves the distinction between observed absence, non-observation, unresolved verification, and unavailable evidence and is intended to support auditability and reproducibility of the reported analyses. The materials are provided for research transparency, methodological audit, and reproducibility. The repository should be interpreted together with the associated manuscript, which defines the corpus protocol, eligibility criteria, coding ontology, provenance hierarchy, deduplication procedure, and analytical limitations.