The principal five-market crawler-access analysis is verified and frozen. This record states what was checked, what agreed, and — equally importantly — which class of reproducibility is not established.
Independent recomputation from frozen source artefacts.
The 6 September 2026 frozen-state inventory independently recomputed the five markets’ principal values from their frozen source artefacts. The reproduced values matched the recorded values with zero discrepancies.
No published numerical error has been identified in the principal C01 access findings.
A file that changes bytes between representations cannot be cited reliably.
Each frozen CSV was compared in three independent representations — the working tree, the Git blob, and the output of git archive — against its recorded SHA256:
worktree SHA256 = Git blob SHA256 = git archive SHA256 = recorded frozen SHA256
Result: 18 of 18 hash comparisons agreed; zero divergence. The governed repository attribute state resolves all six frozen CSVs as text: unset (-text), preserving their literal bytes across checkout and export.
| Market / source | SHA256 |
|---|---|
| Australia | 20d53386bd28cedf87bcc054b91a23b816751d3a47a65215401090d68a513fb1 |
| United States | 669404a4d2aa8bc725d74c7bf2ffacd2d670aa0599a69ee74d4d018155831053 |
| Great Britain | 258cb3750b11e09ac922a65e515d11d70d5c5cc2011d217c031ed01f61b5ff7d |
| Singapore | e5554d62ad1787a0afab5d47544962c0baa08b49203c9bdb232d9154e1da98b4 |
| India — R006-IN crawl | 2494906ec58f59a9260160f825d1af0c809843f1d9bea0b6a5a2bcf435e0f109 |
| India — registry v1.0 | 7d2436a00167ccfefac5ee94214f0ac9d7680343512e8054de8a3c97d4cf5b8d |
The distinction this record exists to keep explicit.
Research reproducibility remains intact. The frozen datasets, the instrument and the recomputation path all hold, and the five market results reproduce from source with zero discrepancies.
Historical publication-layout reproducibility does not. The render implementation used to produce certain historical published artefacts was not captured into repository custody at the time, so those exact published layouts cannot now be regenerated byte-for-byte from the repository alone.
PROVENANCE-DEFECT-002 is a publication-build archival defect. It is not a numerical or measurement defect.
The missing historical render implementation does not invalidate any C01 finding. No figure on any C01 page depends on it. What it affects is the ability to reconstruct a historical document’s exact typeset appearance — a publication-archival property, distinct from the reproducibility of the research itself.
Recorded so the same gap cannot recur silently.
Publication-build source is a governed research artefact. Render implementation, dependency manifest/lockfile, figure-generation source, build instructions and runtime versions must enter repository custody before publication freeze and be versioned alongside instrument, frozen dataset and manuscript.
None of these changes the principal five-market numerical findings.
Carried forward into every C01 publication.
The five-market report may describe observed cross-market differences. It must not claim that geography causes those differences.
C01 measures crawler access only. Crawler access must not be presented as equivalent to AI visibility, AI ranking, AI citation, AI recommendation, digital authority, or downstream commercial performance. Those are separate empirical questions.