How we vet sources
Every reading in a Rabbit Hole curriculum is a real page on the real internet. We check it resolves, that it's free to read, and we say plainly how much weight it deserves. Here's the actual standard.
Credibility tiers
What earns it: Published by a journal, university press, or scientific body with a named editorial board and a stated review process. Government statistical agencies and major research institutions count.
Evidence we require: Name the journal, publisher or institution, and the review process if stated. 'Nature, peer-reviewed' — not 'looks academic'.
What earns it: A masthead, named authors, a corrections policy, and a track record. Wire services, national papers, established magazines, public broadcasters.
Evidence we require: Name the publication and the author. 'Reuters, by [author]' — not 'a news site'.
What earns it: A named author whose credentials in THIS subject are verifiable — an academic, practitioner, museum, or professional body. Reference collections maintained by institutions belong here.
Evidence we require: Name the person or institution and what makes them qualified. 'USHMM, the Holocaust memorial museum' — not 'an expert blog'.
What earns it: Wikipedia, well-moderated forums, collaborative wikis, enthusiast sites with a good reputation. Genuinely useful for orientation; should not be the only source on a contested point.
Evidence we require: Say what the community is and why it's reliable enough to include.
What earns it: No named author, no institution, no clear editorial standard — but the content is on-topic and freely readable.
Evidence we require: Say plainly what could NOT be established. An honest 'unverified' is worth more than a confident wrong label.
Bias is judged separately
A source can be rigorous and still have a point of view — and a neutral source can still be thin. We judge those independently. If something has a political, religious or commercial lean, or it's one side of a contested question, we flag it and say what the leaning is.
Flagged is not rejected. A partisan primary source is often exactly what you want to read — as long as you know what you're holding.
When a topic is thin
Some subjects simply don't have freely readable peer-reviewed work — they're too new, too niche, or the good material is practitioner knowledge rather than published research. When that happens we say so, and we say why.
Knowing the shape of the evidence is part of learning a subject. A learner who knows they're reading one practitioner's experience rather than settled fact reads more carefully — so the warning is a feature, not an apology.
What we check in code
- Every URL is requested by our server before it enters a curriculum. Dead links are dropped, not guessed at.
- Sources are proposed as sites, never as deep links recalled from memory — that's how you get broken citations.
- Each lesson's reading comes from a page we actually found in search results and confirmed loads.
- No two days in a week get the same reading.
Where we're honest about the limits: the tier is assigned by a model applying the standard above, not by a human librarian. It has to cite the specific journal, publisher or author that earned the rating — and when it can't establish provenance, the source is marked unverified rather than given a flattering label.