How far to trust this.

Most are parsed automatically from Western's public Award Search, which lists roughly two thousand awards on one page. A smaller number are curated by hand from faculty and affiliate-college pages that central search does not cover. There are 80 of those, and they are the part you are least likely to find on your own.

We store facts only: names, values, deadlines, and eligibility criteria. The descriptions on each card are written by this tool, not copied from Western.

218 awards are flagged because the parser was not confident, usually because the real rule is written in prose we do not reproduce. They are hidden by default and you can show them from the results page. They are not wrong, just unconfirmed.

1,872 parsed cleanly but have never been read by a person. 12 have. That ratio is the main reason every card links to its official source, and the main reason to use that link.

Some awards state a year of study in prose that we do not turn into a filter. The effect is that you may be shown an award meant for a different year. Annoying, but deliberate: the alternative was guessing, and a wrong guess would hide an award from someone who qualified for it. Everywhere we had that choice, we chose to show too much rather than too little.

Eligibility is decided by fixed rules, never by an AI. The AI only writes the plain-language note explaining a match those rules already made.

Source pages were last checked on 2026-08-16. Links are re-tested and the data re-parsed on a schedule; deadlines in particular drift, so treat the official page as authoritative whenever the two disagree.

Found something wrong? Every card has a “Report an issue” link. Those are read.