Our source hierarchy
Tier 1 — Official labeling and regulators
FDA-approved product labeling published on DailyMed, FDA publications and databases, and published reviews by other national regulators such as the European Medicines Agency. This is as close to a primary source as a patient can read directly, and it is where drug-specific factual claims come from.
Tier 2 — Government consumer health references
MedlinePlus, and health topic pages from the National Institutes of Health institutes, the National Highway Traffic Safety Administration and similar agencies. Written for the public, free of advertising, and appropriate for background on conditions and general safety concepts.
Tier 3 — Professional societies, non-profits, helplines
A small number of sources used for specific, narrow purposes: the American Geriatrics Society Beers Criteria for older-adult prescribing context, MotherToBaby for pregnancy and lactation counselling referral, Poison Control, the 988 Suicide & Crisis Lifeline, and the SAMHSA National Helpline.
What we will not cite
- Blogs and content farms, including large commercial health sites that aggregate without primary sourcing.
- Pharmacy marketing pages, including online pharmacies.
- Forums, social media and review sites. Real experiences, not evidence, and unverifiable.
- Predatory or non-peer-reviewed journals.
- Studies we have not read in full. Citing an abstract, or a description of a study in another article, is not sourcing.
- Sources behind paywalls that a reader cannot check, where a public alternative exists.
- Any source not in our registry.
Why we do not summarise individual studies
Health content routinely cites research in ways that misrepresent it: quoting a finding without its population, treating a small study as settled, repeating a claim traced through three intermediaries back to something that never said it.
Summarising a study responsibly requires reading it in full, understanding its design and limitations, and knowing how it sits within the wider evidence. Where we have not done that work, we do not cite the study — and we do not describe "studies show" in the abstract, which is worse, because it invokes evidence without any way for a reader to check it.
What we do instead is rely on official labeling, which reflects a regulator's assessment of the evidence, and on government consumer resources, which are themselves prepared from appraised evidence. Where the honest position is that we cannot characterise the evidence base, we say so.
How the registry works
Every permitted source is an entry in a controlled registry recording:
- The publisher and the canonical public URL
- Its tier and type
- What the source actually contains
- What it may be used to support
- What it must never be stretched to support
- The date the editorial team last opened it
That fifth item does real work. A general NIH page about anxiety may be cited for background on anxiety as a condition; it may not be cited to support a claim about a specific drug. Recording the boundary in the registry prevents a common failure mode where an authoritative-looking citation is attached to a claim it does not actually support.
Every source we use is listed publicly with links on our sources page.
When links change
Government sites reorganise, and drug labeling is updated as manufacturers submit revisions. Our registry records a last-checked date for each source, and where a source is a search query rather than a fixed document we mark it as such, because results can change.
If you find a broken or redirected link, please tell us through our corrections page.
How broken sources are handled
Source links are checked on a schedule. If an official URL moves, editors update the record and refresh snapshots where available. Temporary outages do not automatically remove a citation, but persistent failures enter a repair queue. Articles that lose a primary Tier 1 source are flagged before major claims remain unsupported.