Measuring Versioned Information Releases and Observable Complementor Links in Generative AI Platform Ecosystems

Generative artificial intelligence (GenAI) model repositories expose versioned information artifacts, but public platform records rarely provide complete adoption histories. This study develops a reproducible architecture for measuring temporal relations between repository documentation (README) revision candidates and observable public complementor links. A task-stratified frame of 3577 models, fixed before downstream outcomes were measured, yielded 365 eligible candidates and 311 matched pairs. Exact model identifiers in public Spaces were traced to repository versions, producing 8137 current model–Space relations and 5316 locatable first appearances. A separately frozen Wave 3 frame of 1440 unused models yielded 62 candidates and 43 pre-event-balanced pairs. This study reports fixed-rule event families, documented matching conditions, full-coverage public link tracing, a linear difference-in-differences specification, and paired-label randomization. Across the primary, enhanced, independent, and pooled designs, confidence intervals include zero and the paired tests do not support stable average differences across pairs. The contribution is an auditable architecture that separates event construction, relationship observability, comparison design, cohort replication, and inference. It does not validate a semantic actionability classifier, measure total adoption, or estimate a causal documentation effect.