Editorial Policy
This site has no human editorial staff, so its rules are engineered rather than aspirational. The core rule is fail-closed publishing: content that cannot pass verification does not get published in softened form — it gets removed, held, or labeled with explicit uncertainty.
Last updated: 2026-08-24.
1. Fail-closed publishing
- Unsupported claims are removed, not weakened. If the fact-check agent cannot tie a claim to a cited source, the claim is deleted from the draft — it is never rewritten as "many parents say…" or "experts believe…".
- Safety-gate failure blocks publication. Safety-critical sections lacking primary-source support hold the entire page back until resolved.
- Three outcomes exist: publish clean, publish with uncertainty clearly stated (UNKNOWN fields visible), or remove the unsupported section and publish the rest. There is no fourth outcome where a guess ships quietly.
- No agent approves its own work. The writer of a claim, the fact-checker of that claim, and the safety gate are separate systems (see AI methodology).
2. Conflicts are stated, not hidden
When official sources disagree — a manual versus a product page, an old versus new revision — pages show the conflict verbatim with both sources and dates, plus the instruction to verify the current manual before use. Conflicts are treated as information for the reader, not an embarrassment to be smoothed into false certainty.
3. No invented experts, testimonials, or experience
| Never published | Published instead |
|---|---|
| "We tested," "during testing," "we measured," "we found it comfortable," "our baby" | "According to the manufacturer…", "The manual specifies…", "The CPSC database shows…", "Based on the verified specifications…" |
| Fake pediatricians, CPSTs, engineers, safety experts, parent reviewers; invented quotes or star ratings from "families" | Direct citation of the actual documentation, with retrieval date |
| "Hands-on review," "long-term tester," "after months of use" | "Document-based analysis of manufacturer-published specifications" |
| Vague authority ("the best pick you can buy") without computed basis | Row-level winners derived from the spec diff, with sources listed on-page |
If no human expert was involved in producing a page, nothing on the page may imply one was. Analysis here is document-based analysis, described as exactly that.
4. Safety language discipline
- Safety guidance is conservative and defers to the product's own manual first, always.
- Where applicable, guidance follows American Academy of Pediatrics recommendations, attributed as such.
- Certification and compliance claims appear only when documented by the manufacturer or regulator — never inferred from category norms.
- Recall information links to the official CPSC/NHTSA notice (recall monitoring) and stays on the page regardless of which product wins the verdict.
- Nothing on this site is medical advice; pages say so explicitly.
5. Affiliate independence
- Verdicts are computed from specifications before any commercial link is placed; commission potential does not enter the diff.
- Affiliate relationships are disclosed on every page that contains them, including when one of two compared products is unaffiliated.
- Products are not excluded from comparison for lack of an affiliate program; unaffiliated retailer links are used where that's the case.
6. Every page must carry data, not just prose
A comparison page earns publication only if it contains something useful beyond generic writing: a verified spec table, a compatibility note, recall status, a total-cost model, a generation-history mapping, or an automated change log. Pages that would be pure word count do not ship.
7. Corrections
Errors found after publication are handled under the dated corrections policy, with a 24-hour target for safety-related errors.