Reconstructed Time: Comment Timestamps in Interface-Level YouTube Extraction: A Measurement Report

yt-dlp emits a Unix timestamp for every YouTube comment it extracts, but the field is not a direct measurement of comment publication time. YouTube’s web interface provides coarse relative-time labels such as “1 year ago”, from which an absolute timestamp is reconstructed at extraction time. This measurement report quantifies the consequences across eleven extractions from seven videos comprising 461,083 comments. In ten of eleven extractions, each distinct relative-time label maps deterministically to a single timestamp value, demonstrating that the timestamp contains no independent temporal information beyond the underlying label. Resolution loss can be severe: in one extraction, 182,784 comments occupy only two distinct timestamp values, with 96.1% of comment pairs exactly tied. Validation against YouTube Data API v3 timestamps on five videos shows that median absolute error ranges from 4.1 days in a three-week-old corpus to 358.3 days in a two-year-old corpus. Two additional failure modes are documented: timestamps can be present when the underlying relative-time label is absent, and extractions crossing midnight UTC can assign different timestamps to identical labels, introducing extraction-order-dependent apparent temporal variation. The report provides simple diagnostic tests and practical recommendations for researchers using interface-level YouTube comment extraction. The contribution is measurement and characterization of a documented approximation, rather than discovery of the approximation itself.