How we research a guide

1. 1. Topic selection. We pick queries owners type when a crate has already failed: escape, bent wire, a hurt mouth, a shepherd that panics when left. The list is reviewed when those searches shift, not on a content calendar detached from the hardware.
2. 2. Source collection. Tier 1 is a statute or a peer-reviewed paper. For floor space that is 9 CFR 3.6. For panic, that is Merck’s behavior chapter, Flannigan and Dodman in JAVMA, and the open reviews we cite. Tier 2 is a specialist page from AVMA, ASPCA, AKC, or Humane World, and we do not use the same Tier 2 URL on more than two guides. The shortlist lives in the authority file.
3. 3. First draft. Anika Singh writes the draft. Every numeric claim has a named source. Desk models, including the wire-beam stress table, are labeled as models with their assumptions, not as laboratory certificates.
4. 4. Editorial review. Naomi Caldwell checks the draft for accuracy and tone. She is not the byline. A claim that cannot be tied to a page we opened gets cut.
5. 5. Update cadence. Each guide shows an updated date and is re-read every six months, or sooner if 9 CFR 3.6 or a cited behavior paper changes. The date on the seed guides is 7 Oct 2026.

Drafting assistance from software is used for research organization and line editing. A person checks every citation against the page, and a second person reviews the guide before it stays up. That human check is the step we will not skip. The declaration in the page source is “human-curated, AI-assisted research.”