How Perplexity ranks sources

Where ChatGPT and Claude add web search as a tool the model invokes when needed, Perplexity is built search-first. Every answer query triggers a live retrieval pass — there's no 'closed-book' mode where Perplexity answers from training data only. The model is Sonar (Perplexity's own family of search-tuned LLMs), and the entire response is grounded in retrieved sources.

Consequence: Perplexity hallucinates less than other engines on factual queries but is also more sensitive to source quality. A query where the retrieved sources are weak produces a noticeably worse Perplexity answer than ChatGPT's.

Three signals are particularly weighted. Recency — Perplexity favours fresh sources more aggressively than ChatGPT or Claude. A 2025 article often outranks a 2023 article on the same topic, even if the 2023 article has higher classical authority. Factual density — pages with specific numbers, dates, named entities, and concrete claims outrank pages with vague prose. Primary-source preference — Perplexity preferentially cites the source that originated a claim over aggregators that quote it. If your data was first published on your own site, Perplexity tends to cite you over the news outlet that picked it up.

Perplexity Pro Search runs a more elaborate retrieval pass: more queries, more sources, longer answers, with the Sonar model doing more reasoning per answer. Citation patterns shift — Pro Search tends to surface more obscure but high-relevance sources, while free Perplexity sticks closer to top-ranked results. If you're optimising for Perplexity visibility, audit both free and Pro Search separately — they can differ materially on the same query.

Four high-leverage moves. (1) Lead with concrete facts — numbers, dates, specific claims. Perplexity rewards factual density. (2) Keep dateModified aggressively current — Perplexity weights recency more than other engines. Even minor revisions every 60–90 days help. (3) Publish primary data — original research, surveys, internal benchmarks. Perplexity's primary-source preference means original data tends to get cited even on hot topics where many secondary sources exist. (4) Add Article schema with explicit author + publisher entities — Perplexity's source-quality ranking takes named authorship seriously.