How do AI assistants like Perplexity choose their sources?

Answer engines run a retrieval pipeline: they search the web for candidate pages, read the top results, and synthesize an answer citing the passages that most directly and credibly resolve the query. Selection happens at the passage level, not the site level.

That has two consequences. Ranking still matters — you must be in the candidate set to be quoted. And extractability decides the winner among candidates: the page that states the answer cleanly beats the higher-authority page that buries it.

Cross-source corroboration is the tiebreaker: models favor claims confirmed by multiple independent surfaces, which is why consistent facts across your site, profiles, and third-party mentions quietly compound.