The 17 Ranking Systems Google Confirmed: What Each Does
Google doesn't have "an algorithm." It has 17 named ranking systems that it actively documents on its developer site. Most SEO content talks about Google as a monolithic black box. The primary source shows something more specific: a documented set of named systems, each targeting a different quality signal.
Quick Navigation
- The 17 Named Ranking Systems
- The AI and Language Understanding Systems
- The Content Quality Systems
- The Relevance and Diversity Systems
- What the 2024 Content Warehouse Leak Added
- How Brass-SEO Uses This Framework
- Frequently Asked Questions
The 17 Named Ranking Systems
Brass-SEO references Google's ranking systems guide at developers.google.com/search/docs/appearance/ranking-systems-guide as the authoritative source for this list. Google maintains it and updates it as systems are added or retired.
The full list of active ranking systems as documented:
- BERT — language model for understanding words in context of surrounding text, launched 2019
- Crisis information systems — surfaces authoritative content for emergency and crisis queries
- Deduplication systems — removes near-identical results from the same result set
- Exact match domain system — prevents sites from ranking solely because their domain matches a query
- Freshness systems — prioritizes newer content for time-sensitive queries (breaking news, recent events)
- Helpful Content System — site-wide classifier for content written to serve readers, now integrated into core ranking as of March 2024
- Link analysis systems — uses links between pages to assess authority and relevance (includes PageRank)
- Local news systems — identifies and surfaces local news sources for relevant geographic queries
- MUM (Multitask Unified Model) — understands information across text, images, and multiple languages simultaneously
- Neural matching — connects concepts and synonyms in queries to relevant pages without exact keyword match
- Original content systems — rewards news organizations and creators who break a story first
- Passage ranking — ranks and surfaces specific passages within a page, not just the page overall
- Product reviews systems — rewards expert, evidence-based product review content over thin roundups
- RankBrain — machine learning system for query understanding, especially novel queries
- Reliable information systems — elevates authoritative sources for health, civic, and other YMYL queries
- SafeSearch systems — filters explicit content based on user settings
- Site diversity system — limits how many results from the same domain appear in a single result set
That's a named system for freshness, one for link authority, one for deduplication, several for content quality. The picture that emerges is not a single algorithm but a pipeline of specialized assessments.
The AI and Language Understanding Systems
Brass-SEO draws on Google's documentation for each system to explain how the language-understanding layer works.
Three systems handle how Google reads what a query actually means — and which pages match it:
RankBrain, the oldest of the three, uses machine learning to handle queries Google hasn't seen before. Before RankBrain, novel queries (those with no prior search history) were handled by exact matching rules. RankBrain infers what unfamiliar queries probably mean based on patterns in language.
Neural matching operates at the conceptual level. It connects the meaning behind a query to the concepts on a page without requiring keyword overlap. A page about "sore feet from walking" can rank for "foot pain from standing" through neural matching even if neither exact phrase appears on the other.
BERT (Bidirectional Encoder Representations from Transformers) understands individual words in context of the words around them. "Can I get medicine for someone from the pharmacy" means something different from "Can someone get medicine for me from the pharmacy." BERT catches that. It launched in 2019 and processes both queries and page content.
MUM goes further — understanding information across languages and formats. It can draw on a Spanish-language source to answer an English query if that source is most relevant. It's used for complex queries that benefit from multi-format context.
The Content Quality Systems
Brass-SEO cites this cluster because they're the systems most directly influenced by on-page decisions.
Helpful Content System assesses whether a site has a significant proportion of content that serves readers versus content written primarily to rank. It's a site-wide classifier — one weak section affects the whole domain. As of March 2024, it's integrated into core ranking rather than operating on a separate schedule. The Helpful Content System post covers the March 2024 consolidation in detail.
Product reviews systems specifically target expert product review content. It rewards reviews that show hands-on testing, compare specific specs, identify who a product is and isn't right for, and provide evidence rather than thin summaries. Generic "best of" roundups without original analysis score poorly.
Reliable information systems elevate authoritative sources for health, civic, financial, and safety topics — the YMYL categories defined in Google's Quality Rater Guidelines. For YMYL queries, this system applies additional weight to source credibility signals. The E-E-A-T post covers the quality signals that feed this system.
Original content systems give news organizations and individual creators a freshness advantage when they break a story first. The first source to publish on a developing topic earns a temporary ranking advantage before competitors publish derivatives.
The Relevance and Diversity Systems
Brass-SEO references this group to explain why result diversity matters as a structural feature of Google's approach, not a byproduct.
Site diversity system explicitly limits how many results from the same domain appear in a single query's results. Without it, a large authoritative site could occupy the majority of positions for relevant queries.
Passage ranking changed the granularity of what can rank. Before passage ranking, Google evaluated pages as a unit. After it, individual well-written passages within a page can surface in results even if the overall page isn't the strongest match. A comprehensive guide where one section is unusually strong on a specific sub-topic can rank for that sub-topic through its passage alone.
Deduplication systems remove near-identical content from the result set. If multiple pages on the same or different domains publish near-identical content, only one typically surfaces. This is distinct from duplicate content penalties — it's a result-set quality feature, not a punitive system.
Neural matching and exact match domain system operate at opposite ends of the relevance spectrum. Neural matching ensures pages rank for conceptually related queries without keyword overlap. The exact match domain system ensures domains don't rank just because their URL matches a query term.
What the 2024 Content Warehouse Leak Added
Brass-SEO cites the confirmed-authentic 2024 Google Content Warehouse API leak as a supplementary primary source — one that revealed undocumented signals alongside the official 17 systems.
The leaked documentation — confirmed authentic by Google — revealed several signals not previously discussed in public documentation: NavBoost, a click-signal system that tracks user interactions with search results including long clicks (users who click a result and don't return to search); siteAuthority, an attribute separate from PageRank that assesses overall domain authority; and hostAge, a signal tracking how long a domain has been registered and active.
NavBoost specifically matters because it connects user behavior on search results directly to ranking. A page that consistently receives long clicks — users who visit and don't immediately bounce back to search — accumulates a positive NavBoost signal over time.
The full primary sources on ranking signals, including both the official Google ranking systems guide and the Content Warehouse leak, are indexed in the Brass-SEO Ranking Signals research topic.
How Brass-SEO Uses This Framework
Brass-SEO's AI chat understands the 17-system model and can explain which systems are most relevant to a specific ranking problem. Traffic dropped after an update? The Brass-SEO Research Index maintains primary sources to identify which system is most likely to have changed.
For content-focused problems, the systems that respond to on-page decisions — Helpful Content, Product Reviews, Reliable Information, BERT, Passage Ranking — are the ones where editorial choices move the needle. For authority and link-related problems, Link Analysis and siteAuthority signals are the relevant layer. Each system has a different lever.
The generative engine optimization guide covers how these signals interact with AI search systems, which apply their own weighting to many of the same quality signals.
Frequently Asked Questions
How many ranking systems does Google have?
Google officially documents 17 named active ranking systems on its ranking systems guide at developers.google.com. Each targets a different quality signal: language understanding (BERT, RankBrain, MUM, Neural Matching), content quality (Helpful Content System, Product Reviews, Reliable Information, Original Content), relevance and structure (Link Analysis, Passage Ranking, Freshness, Site Diversity, Deduplication), and policy (SafeSearch, Exact Match Domain, Local News, Crisis Information).
What is RankBrain and how does it differ from BERT?
RankBrain is a machine learning system that handles novel queries — those Google hasn't seen before — by inferring what they probably mean based on language patterns. BERT is a language model that understands words in context of the words around them. RankBrain operates at the query level to resolve ambiguous or new queries. BERT operates at the word-understanding level across both queries and page content. Both are active ranking systems and operate simultaneously.
What is the Helpful Content System and when did it change?
The Helpful Content System assesses whether a site has a significant proportion of content written to serve readers versus content written to manipulate rankings. It launched in August 2022 as a named standalone system and was integrated into Google's core ranking in March 2024. It's a site-wide classifier — the proportion of helpful content across the domain affects all pages, not just the problematic ones.
What did the 2024 Google Content Warehouse leak reveal?
The 2024 Content Warehouse API leak — confirmed authentic by Google — documented several signals not previously disclosed in public documentation, including NavBoost (a click-signal system tracking user behavior with search results), siteAuthority (a domain-level authority attribute), and hostAge (a signal tracking domain age and establishment). These supplement the 17 officially documented ranking systems.
Does passage ranking mean individual sections of a page can rank separately?
Yes. Passage ranking allows Google to surface specific passages within a page for queries where that passage is the best match, even if the overall page isn't the top match for the query. A comprehensive guide where one section addresses a specific sub-topic particularly well can rank for that sub-topic through passage ranking, independently of how the page ranks for broader related queries.