When several pages could answer the same question, AI engines choose sources on a handful of recurring factors: how precisely the page answers the exact question, how easy the answer is to extract, how fresh and internally consistent the facts are, how authoritative the domain looks for the topic, and whether other sources corroborate the claims. No single factor wins; the cited page is usually strong on most of them at once.
The selection factors, ranked roughly
| Factor | What the engine checks | Your lever |
|---|---|---|
| Relevance | Does the page answer this exact question? | One page per question, question in the heading |
| Extractability | Is there a clean, self-contained answer? | Answer-first paragraphs, lists, tables |
| Freshness | Are facts and dates current? | Refresh cadence, honest dateModified |
| Authority | Is this source credible for this topic? | Topical depth, entity clarity, mentions |
| Corroboration | Do other sources agree? | Consistent facts everywhere you appear |
How ties actually break
Between two credible, extractable pages, engines tend to prefer the one that is more recent, more specific, and better corroborated by independent sources. That is why a smaller site with a precise, current, well-structured answer regularly beats a bigger brand's broad, aging guide, and why the same question can cite different sources in different engines, each weighting the factors differently.
What to do with this
- For every priority question, check who is cited today and diagnose which factor they win on.
- Match their strength (structure, freshness) before trying to out-authority them.
- Fix contradictions between your pages, self-disagreement is the cheapest factor to repair.
- Publish something they cannot copy: your own data, numbers or verdicts.
