TRACE-S: provenance and traceability of bibliographic searches in health research—methodological specification and exploratory documentary application

Objective: To specify Traceability in Scientific Search (TRACE-S) and examine its exploratory application to documenting the provenance of bibliographic searches. Methods: The specification was applied to three PubMed query versions concerning artificial intelligence and periapical lesions. V01 and V02 did not restrict search fields; V03 limited the search to title and abstract (Title/Abstract) and was repeated by the same agent. No date, language, species, or study-design filters were applied. Four executions were reconciled by PubMed identifier (PMID). The 20 lowest-numbered PMIDs were examined using artificial intelligence-assisted extraction followed by human verification. The internal integrity of the records was subsequently checked. Results: TRACE-S comprises six operational stages, seven documentary categories, and the Scientific Search Quality Checklist (SSQC-15), an unvalidated preliminary checklist. The four executions retrieved 123, 134, 127, and 127 records, corresponding to 134 unique PMIDs and 511 execution–document associations. Of the 20 records examined, seven were operationally included, eleven were excluded, and two remained unresolved; five inclusions relied on contextual inference. The audit confirmed the internal consistency of the retrieved sets, documentary links, and recorded history. Conclusion: The preserved dataset showed reconciliation and internal traceability across queries, executions, records, and eligibility judgments. The validity of the judgments, independent reconstruction, superiority over existing approaches, and validity of the SSQC-15 were not demonstrated.