Google doesn't run a single algorithm - it runs a stack of interconnected systems, each evaluating different aspects of a page or query.
PageRank. Launched with Google in 1998, PageRank was the original link-analysis algorithm built by Larry Page and Sergey Brin at Stanford. It treats every link as a citation, scoring a page by the number and authority of the pages that point to it. More than 25 years later it still anchors Google's ranking stack. Google retired the public Toolbar PageRank score in 2016, but link-based authority remains a core signal.
Hilltop. Developed by Krishna Bharat in the late 1990s and adopted by Google in the early 2000s, Hilltop refined PageRank with the idea of topical authority. Rather than counting all links equally, it gives more weight to links from expert pages on the same subject than to links from unrelated sites. The effect was to push ranking away from raw link volume and toward subject relevance. That principle, on-topic authority over generic popularity, still shapes how Google evaluates links.
RankBrain. Introduced in 2015, RankBrain was Google's first major use of machine learning in ranking. It was built to interpret the roughly 15 percent of daily searches Google had never seen before, mapping unfamiliar or ambiguous queries to similar known ones. Google has described it as one of its three most important ranking signals. RankBrain marked the shift from matching keywords to interpreting intent.
BERT & MUM. BERT arrived in 2019 as a language model that reads a query bidirectionally, grasping how word order and small words like 'to' or 'for' change meaning. It initially affected about one in ten English searches before expanding across languages. MUM followed in 2021, described by Google as 1,000 times more powerful than BERT, adding multimodal and multilingual understanding across 75 languages. Together they moved Search from keyword matching to genuine language comprehension.
Helpful Content System. Launched in August 2022, the Helpful Content System is a sitewide signal that demotes content created primarily to rank rather than to help people. Because it evaluates a domain as a whole, a thin or search-first section can weigh down stronger pages around it. In the March 2024 core update, Google folded it into its core ranking systems and retired the standalone classifier. Helpfulness is now assessed inside every core update rather than as a separate process.
Penguin & SpamBrain. Penguin launched in April 2012 to target manipulative link schemes, paid links, and over-optimized anchor text. In 2016, Penguin 4.0 became real-time and part of the core algorithm, devaluing bad links rather than penalizing whole sites. SpamBrain, Google's AI-based spam-prevention system, has run since 2018 and was detailed publicly in 2022. The December 2022 link spam update used SpamBrain to detect and neutralize spammy links at scale, making spam defense continuous and machine-driven.