How Perplexity ranks content: Research uncovers core ranking factors and systems

Want to know how content is scored, ranked, and in some cases, discarded by Perplexity? Independent researcher Metehan Yesilyurt analyzed browser-level interactions with Perplexity’s infrastructure to reveal how the AI answer engine evaluates and ranks content.

Why we care. Everybody involved with driving SEO and/or GEO success wants to understand how to gain visibility (citations and mentions) in AI answer engines. This research (albeit unverified at this point) offers some clues about Perplexity’s ranking signals, manual overrides, and content evaluation systems that could improve your optimization strategies for Perplexity (and possibly other answer engines) to gain a ranking advantage.

Entity search reranking system. One significant Perplexity system uncovered is a three-layer (L3) machine learning reranker. It is used for entity searches (people, companies, topics, concepts). Here’s how it works:

Initial results are retrieved and scored, like traditional search.
Then, L3 kicks in, applying stricter machine learning filters.
If too few results meet the threshold, the entire result set is scrapped.

This means quality signals and topical authority are super important for L3 – and keyword optimization isn’t enough, according to Yesilyurt.

Authoritative domains. Yesilyurt also discovered manual lists of authoritative domains (e.g., Amazon, GitHub, LinkedIn, Coursera). Yesilyurt wrote:

“This manual curation means that content associated with or referenced by these domains receives inherent authority boosts. The implication is clear: building relationships with these platforms or creating content that naturally incorporates their data provides algorithmic advantages.”

YouTube synchronization = ranking boost. Another interesting find: YouTube titles that exactly match Perplexity trending queries see enhanced visibility on both platforms.

This hints at cross-platform validation. Perplexity might validate trending interest using YouTube behavior – rewarding creators who act fast on emerging topics, according to Yesilyurt.

Core ranking factors. Yesilyurt documented dozens of what he called Perplexity’s “core ranking factors” that influence content visibility:

New post performance: Early clicks determine long-term visibility.
Topic classification: Tech, AI, and science get boosted; sports and entertainment get suppressed.
Time decay: Publish and update content frequently to avoid rapid visibility declines.
Semantic relevance: Content must be rich and comprehensive – not just keyword-matched.
User engagement: Clicks and historic engagement signals feed performance models.
Memory networks: Interlinked content clusters rank better together.
Feed distribution: Visibility in feeds is tightly controlled via cache limits and freshness timers.
Negative signals: User feedback and redundancy checks can bury underperforming content.

What’s next. Yesilyurt said success on Perplexity requires a combination of strategic topic selection, early user engagement, interconnected value, continuous optimization, and prioritizing quality over gaming.