Sample findings report — Scholar Forensics

1. Verdict

Citation metadata on the new platform no longer matches the PDFs, and Google Scholar dropped articles after the move. Redirects were correct and are not the cause. Confidence: High for the metadata mismatch, Medium for it being the only cause.

2. Evidence

ID	Observation	Source	Time (UTC)
E1	Search Console shows indexed pages unchanged after the move; Scholar coverage of 40 sampled titles fell from 38 to 9.	GSC export; dated Scholar baseline	2026-01-12 / 2026-02-09
E2	All 40 sampled old URLs return one 301 to the matching new URL.	Redirect crawl	2026-02-10 09:14
E3	On 31 of 40 sampled articles, citation_title in the page differs from the title printed in the PDF.	Page tags vs PDF text	2026-02-10 11:02
E4	citation_pdf_url returns an HTML viewer page, not a PDF, on 22 of 40.	Direct fetch	2026-02-10 11:40
E5	The platform template changed on the move date; the old template emitted both tags correctly.	Client changelog	2026-01-08

3. Confidence, defined

High: the cause was reproduced and confirmed by an independent source.
Medium: consistent with several facts but not reproduced.
Low: a hypothesis; more data is needed.

4. Not established

Whether Scholar also penalised the journal for the 3-week gap in full-text availability (E4). The evidence does not separate the two.
Coverage of articles older than 2022, which were not sampled.

5. Remediation, in order of risk

Restore citation_pdf_url to a direct PDF address (E4). Staging first; no DOI target changes.
Regenerate citation_title from the same source as the PDF (E3). Test on 10 articles, then all.
Re-submit the sitemap; leave the redirects in place for at least a year.

6. Watch plan

Repeat the 40-title Scholar baseline at 2, 6 and 12 weeks after the fix. Report the movement against the first baseline, not against the last check.