Evidence
Every note this tool raises comes with the guidance clause behind it. This page shows the research behind the guidance: the studies that found harm, the studies that found protection, and the studies that found the picture is murkier than a headline. Each record is written from the published abstract and checked against Europe PMC. Nothing on this page was generated by a model.
These records live in a second Google Agent Search data store. On every review, the records that bear on a note's trigger classes are retrieved in code and attached to the note as evidence. The corpus is versioned like the guidance corpus and served from GET /v1/evidence.
Werther effect: imitation after prominent or detailed portrayals, strongest when a method is shown. Papageno effect: the protective association when a story shows someone coming through a crisis. Both are real findings; the field argues about size, mechanism, and how far results from news transfer to fiction. We show that argument rather than settle it.
That any single scene causes or prevents harm. That following guidance makes a script safe. That this evidence base is complete. The tool reports what published guidance says and what research has found; the writer decides, and a human adviser remains the right escalation.
Method
Peer-reviewed studies and systematic reviews on media and suicide, with emphasis on fictional and film portrayals. Each has a DOI and, where indexed, a PubMed identifier.
Titles, authors, journals, years, and identifiers were checked against the Europe PMC record on the date stated. Findings are paraphrased from the abstract with the numbers the abstract reports.
Each record is tagged with the trigger classes it bears on, so the right research appears beside the right note. Disagreeing studies are tagged to the same classes on purpose.