• by Laurie Sullivan , Staff Writer, Yesterday

Metehan Yesilyurt, growth marketing manager at AppSamurai, analyzed browser-level interactions with Perplexity’s infrastructure and believes he has discovered how content is evaluated and ranked.

It appears the AI engine follows a three-layer ranking system for the large language model (LLM), and encodes request patterns that reveal critical insights about how it prioritizes content. He says Perplexity’s ranking algorithms explain how some content has ranking advantages.

“The browser request layer contains additional signals that aren’t visible through standard API interactions,” Yesilyurt wrote in a post.

Content creators who understand these request patterns should align optimization strategies with deeper infrastructure requirements, he wrote.

He said the reranking system explains why some well-optimized content fails to appear for searches. While it may rank well initially, it could fail the LLM’s quality assessment. Success requires not just keyword optimization but topical authority and quality signals that satisfy machine learning.

He pointed to "new_post_impression_threshold" as one of Perplexity’s most critical factors in its ranking algorithm.

“When content is published, it enters a crucial window defined by new_post_published_time_threshold_minutes where performance metrics determine long-term visibility,” he wrote, adding that posts must achieve specific engagement levels during this window before the algorithm will amplify the content.

Manually configured authoritative domains, he felt, is another significant discovery in Perplexity’s ranking system. It appears that Perplexity maintains curated lists of high-trust sources across different categories.

The full list of exact prompts was not shared in his post, but marketers should pay attention to quality signals and topical authoritative sites, he wrote. Keyword optimization isn’t enough.

Similar to optimizing in traditional search, authoritative domains include Amazon, eBay, Walmart, Best Buy, Target, Costco, Github, Stack, Facebook, LinkedIn, Reddit, Pinterest, Booking, Kayak, among many others.

Yesilyurt experimented with the approach he outlined in his blog post. He had success, but did mention when it fails to index properly in Perplexity when the content fails to indexed by Google.

Last weekend, Yesilyurt ran an experiment and managed to get his website to trend in Perplexity’s AI search engine. While he hit thresholds in his blog’s metrics for some queries and topics, he confirmed that overall what he discovered worked.

YouTube titles that match Perplexity trending queries exactly see better visibility, but the dozens of Perplexity’s “core ranking factors” became the real find. Things like what influences content visibility and how early clicks on a piece of content determine long-term visibility. Content also must have semantic relevance and not just keyword matching. There’s a long list in the blog post.