How mensresearch.ai indexes, ranks, and synthesizes peer-reviewed research on men's health.
mensresearch.ai aggregates peer-reviewed research from PubMed (36 million+ biomedical citations) and Semantic Scholar (220 million+ scholarly papers across disciplines), filtered to men's health relevance via MeSH term matching and embedding similarity to a curated seed corpus. As of publication, 1.3M papers are indexed.
Preprints and non-peer-reviewed sources are explicitly excluded.
For each search query, we compute a blended relevance score:
Score = 0.7 × semantic relevance + 0.2 × log-normalized citations + 0.1 × recency decay
Papers are then quality-filtered: abstracts must exceed 100 characters, be published ≥ 2015 or carry citations, and be ≥ 80% ASCII (filters OCR noise and non-English junk).
The top 10 quality-filtered sources for a query are sent to Claude Sonnet 4.6 (Anthropic) which returns a structured JSON response with four components:
If fewer than 3 quality sources are available, the system returns "insufficient evidence" rather than synthesizing on weak ground.
See our Medical Disclaimer for details.
Algorithm v1.0 went live 2026-04-22. Material changes — weighting tweaks, prompt updates, filter additions — are logged at /methodology/history (coming soon).
Editorial questions or corrections: [email protected].