This page openly explains how our content is produced, reviewed, and risk-graded, and what role AI plays. Our goals are verifiable, correctable, and never overstated.

How content is produced

• Each answer's conclusion comes from a structured knowledge base and authoritative sources first; AI then drafts the natural-language wording.
• AI does not decide the "can you eat it" conclusion — it only turns verified facts into clear, consistent wording.
• Before publishing, pages pass automated checks (banned phrasing, structure, disclaimers) and a human review process.

AI use disclosure

• AI is used only to help draft and translate wording, at a low temperature setting for stability.
• When reliable evidence is missing, AI is instructed to flag "needs a source" rather than fabricate a conclusion.
• High-risk health content is never auto-published on AI output alone.

Risk grading & review gates

• Content is graded green, amber, red, or black; high-risk topics (pregnancy, chronic disease, infants, medication, poisoning, first aid) are not auto-published.
• People's-health (YMYL) content is marked with a review status and only becomes a public conclusion after human or expert review.
• Every page shows its review status, last-updated date, and who reviewed it, so claims can be verified.

How we choose sources

• We prioritize government bodies, international organizations, and authoritative professional institutions (e.g., WHO, FDA, ASPCA, Merck Veterinary Manual, Chinese Nutrition Society, CFSA).
• Sources are used to cross-check conclusions and risk boundaries, and pages label the institution and authority tier.
• When a source is broken or outdated, it is replaced through the correction and update process.

Found a factual error, an outdated source, or an inappropriate risk level? Submit it via the correction form; we will review and update it.