Graded at ingestion, not at answer time
Every passage carries its credibility grade from the moment it enters the corpus. Grading is a property of the source, so it cannot be adjusted to make a particular answer look better supported.
Knowledge
Corpus claims are generated from the database rather than written into copy, so what you read here is what the retrieval layer actually holds.
Evidence before eloquence
OpenSupport prioritises authoritative, peer-reviewed and institutional material. Lower-weight commercial content is labelled and never serves as the sole basis for an answer.
7,500+
sources indexed across 25+ therapeutic areas
11,000+ passages are retrievable today. The corpus is refreshed continuously, and stale content is archived rather than served.
Share of 11,000+ retrievable passages as of September 2026. 0 rows without a usable embedding are excluded — they cannot be retrieved, so they are not counted.
How the corpus is governed
Every passage carries its credibility grade from the moment it enters the corpus. Grading is a property of the source, so it cannot be adjusted to make a particular answer look better supported.
Manufacturer-published material is retrievable and clearly labelled, but it sits at 0.2% of retrievable passages and cannot carry an answer on its own.
Sources are re-crawled on a schedule matched to how fast each one changes — regulatory labels more often than reference material. Content that goes stale is archived rather than served.
0 rows currently have no usable embedding, which makes them invisible to vector search. They are excluded from every number published on this site.
Figures derived 2026-09-14 and regenerated from the corpus on each release. Exact values: 7,620 sources, 11,383 retrievable passages, 25 therapeutic areas. Published claims are rounded down so they remain true between regenerations.
The corpus is the product. The quickest way to judge it is to probe the edges of your therapeutic area.