Preprint — not yet peer reviewed. v3.0 (2026-07-15): adds controlled evidence on self-owned accounts. The nickname-rewrite intervention is now replicated on a second, independently created account; a within-account nickname reversal shows the rewrite is repeatable and reversible (a controlled three-value chain over the same sounds); an anonymous read repeated through VPN exit nodes confirms the value is identical across TikTok's US, EU and ROW serving regions; and an account-visibility change shows that making an account private closes the live channel only prospectively, without erasing third-party archives or offline records. The reproducibility deposit now includes a per-item table for the high-volume account, so its self/reuse partition and offset analysis are independently recomputable. Bibliographic corrections included. v2.0 (2026-07-12): corrected the account-identifier discussion of v1.0 (the high bits of the TikTok userId do not reliably encode the account's creation time; all creation-date claims were removed and the userId is used solely as an immutable anchor). No result was affected. We document an identity channel not previously reported in TikTok forensic account attribution: for original sounds (music.original=true), the authorName field resolves at query time to the current display name of the creating account, while music.id and the immutable userId stay fixed. Renaming the account retroactively rewrites the exposed name of all its original sounds, leaving a reconstructible chain of (UID, authorName, date) triples. Validated entirely on public accounts (seven accounts, 25 sounds time-anchored via the Wayback Machine) plus sealed single-variable controlled experiments on two self-owned accounts. Reproducibility package (published): DOI 10.5281/zenodo.21382360 (CC BY 4.0).
Luis Ángel Mercado Hidalgo (Wed,) studied this question.