How Perplexity's leaked ranking signals are rewriting SEO

Executive overview

Perplexity's ranking signals were leaked, and they bear little resemblance to Google's. The platform runs its own discovery system that evaluates content the way a knowledgeable person evaluates advice — looking for logical connections, consistent depth, and validated authority across sources.

Keyword optimisation is no longer sufficient. Visibility on AI search engines now requires building topical authority across platforms, keeping content fresh, and earning entity recognition from credible publications.

The core shift: AI systems reward genuine expertise that compounds over time, not isolated pages optimised for search terms.

Perplexity traffic quality vs ChatGPT

• Study of 32 companies (each $100M+ revenue) compared Perplexity and ChatGPT traffic
• ChatGPT sends more raw visitors; Perplexity sends fewer but higher-quality ones
• Perplexity visitors spend more time on site, view more pages per session, convert at higher rates
• The relevant metric shifts from volume to intent quality

The discovery feed changes how content is found

• Perplexity has a discovery tab alongside its search feed — surfaces content users didn't query
• Analogous to TikTok's For You page: proactive, not reactive
• Early engagement signals (clicks, time on page in the first hour) determine whether content rises or fades
• Treat every publish like a launch event: pre-build anticipation and drive immediate engagement

Content half-life and freshness signals

• All content has a built-in decay rate; stale content is deprioritised automatically
• Decay speed varies by topic: educational and strategic content holds longer than fast-moving news
• Wikipedia ranks consistently because its pages are continuously updated — the system reads updates as proof of ongoing relevance
• Each content update signals mastery, not just maintenance

Topic clusters and cross-platform authority

• Perplexity uses memory networks — a clustering system where related content pieces reinforce each other
• Connected posts (e.g., a beginner lifting series across three articles) let the AI pull from all pieces to deliver a complete answer
• Scattered, unrelated content signals no clear expertise — the AI doesn't treat you as a specialist
• YouTube titles influence blog post rankings; LinkedIn articles reinforce website content
• Perplexity leans on Reddit and YouTube; ChatGPT leans on Wikipedia and Reddit; Google AI Overviews favour YouTube and Quora
• Content clusters confined to one platform miss the ecosystems AI systems already trust

Building entity recognition and semantic authority

• Entity recognition: AI maps recognised people and brands as credible authorities in specific fields
• Consistent citation in respected publications (Forbes, TechCrunch, industry outlets) causes the AI to associate your name with a subject
• Digital PR is now a core AI SEO strategy — 28% of 200 surveyed companies are already increasing digital PR budgets for AI visibility
• A citation from a credible publication signals entity validation, not just a backlink
• Semantic coverage: the AI evaluates how completely you address a topic across all its angles
• Multi-format presence (video, blog, podcast, social) expands authority footprint
• Depth in three subjects outperforms shallow coverage across thirty

What to stop doing

• Optimising for keywords people type in, ignoring the discovery feed
• Publishing once and assuming the content will remain visible
• Scattering content randomly across unrelated topics
• Staying on one platform only
• Producing repetitive, shallow posts — high bounce rates and short time-on-page are counted against you