Neural matching is an AI system Google uses to understand the concepts represented in queries and pages and match them to one another. Google introduced it in 2018, confirms it in its ranking systems guide, and describes it as a critical part of how it retrieves candidate documents. It means a page can be found for searches phrased differently from the page. Sites cannot optimise for it directly, but pages written only in internal jargon give it less to work with.
What is neural matching?
Neural matching is an AI system Google uses to represent the concepts in a query and in a page, and to match them, even when they share few words. Google introduced it to Search in 2018.
Neural matching is an AI system that Google uses to understand representations of concepts in queries and pages and match them to one another.
Google, A guide to Google Search ranking systems, Google Search Central
How does neural matching differ from RankBrain?
Google positions neural matching as a retrieval system: it helps choose which documents from the index are candidates for a query. RankBrain helps rank them. In IR terms, neural matching is closer to semantic retrieval than to reranking.
In March 2019 Google's Search Liaison account summarised the split as RankBrain relating pages to concepts and neural matching relating words to searches, and called neural matching a kind of super-synonym system.
This is what makes neural matching such a critical part of how we retrieve relevant documents from a massive and constantly changing information stream.
Pandu Nayak, Google, How AI powers great search results, Google Blog
What evidence is there that Google uses neural matching?
The evidence label is Confirmed. Google lists neural matching in its ranking systems guide. Google's Danny Sullivan said in September 2018 that it was affecting 30% of queries.
Google has also said what it is not. When SEOs linked it to the March 2019 core update, Google said neural matching was unrelated to core updates, and John Mueller described separating out one machine learning component as "more artificial than really useful".
For example, neural matching helps us understand that a search for "why does my TV look strange" is related to the concept of "the soap opera effect."
Google Search Liaison, Google Explains Neural Matching vs RankBrain, Search Engine Roundtable
What should sites do about neural matching?
Neural matching exists to bridge the gap between how authors write and how searchers ask. Google says it works without any special effort, but it has to match a concept that is actually described on the page.
The failure an audit can find is a page that describes its core concept only in internal, product or brand terminology. If a page calls a feature by a trademarked name and never says in plain words what it does, there is less for any concept matching system to connect to the searcher's plain language query. Add a plain description; do not add keyword lists.
Which Laurelin audit checks test for neural matching?
Neural matching runs inside Google's retrieval and cannot be observed from a site. These checks are proxies for whether a page describes its concept in terms a searcher would recognise.
Core concept described only in jargon or brand terms: the new check for this page.
Commodity content: concept matching helps find a page, not make it worth ranking.
Content not matching search intent: a matched concept with the wrong intent still fails.
Title is just brand on a content page: a brand only title gives no concept signal.
Keyword stuffing and over-optimisation: synonym lists are the wrong response to concept matching.
What are the key dates for neural matching?
Frequently asked questions about neural matching
Is neural matching the same as RankBrain?
No. Google describes neural matching as mainly a retrieval system that relates words to searches, and RankBrain as a ranking system that relates pages to concepts.
Did neural matching cause a core update?
Google said in 2019 that neural matching was not connected to the March 2019 core update or earlier updates.