How to Optimize for RankBrain and Semantic Search in 2017

RankBrain and Semantic Search has sparked a lot of buzz in the SEO space. We've heard about entity SEO, conversational content, optimizi...

RankBrain and Semantic Search has sparked a lot of buzz in the SEO space. We've heard about entity SEO, conversational content, optimizing for topics (as opposed to keywords), and even completely ditching old-school SEO tactics, like link building and keyword targeting, in favor of creating the most relevant and useful piece of content there is, and letting Google do the rest.

But is Google really giving up on keywords, and should SEOs do the same? What exactly does "optimizing for relevance" mean, how do you do it, and can you rely on it alone? How, after all, does semantic search work, and where do you get started? This article is an attempt to answer these questions.

What's Semantic Search?

Semantic search aims to improve search accuracy by understanding searcher intent, contextual meaning of terms, and relationships between words to provide more relevant search results.

It all started with Google's Hummingbird update back in 2013. Hummingbird uses context and searcher intent (as opposed to individual keywords in a query) to ensure that "pages matching the meaning do better, rather than pages matching just a few words".

RankBrain (launched in October 2015) forms part of Google's Hummingbird algorithm. Its purpose is similar to that of Hummingbird, but the mechanics behind it are different. RankBrain is a machine learning system that includes two components: the query analysis part and the ranking part. For the former, RankBrain attempts to better interpret queries (particularly the rare or unfamiliar long-tails) by associating them with more common queries, so that it can provide better search results in response.

RankBrain's ranking component analyzes the pages in Google's index and looks for specific features (e.g., usage of certain related terms) that make those pages a good fit for the query. Such "features" are determined by analyzing the best-performing search results (according to Google's user satisfaction metrics, such as SERP click-through rate, pogo-sticking, time on page, etc.) and looking for similarities between these pages. As a result, the pages that are deemed to be good responses to the query may not even contain the exact words from the query, but are nonetheless relevant. Google's recently said that RankBrain is "involved in every query", and affects the actual rankings "probably not in every query but in a lot of queries".

There are many aspects to pay attention to if you're looking to embrace semantic search, from choosing what to focus your pages on to researching keywords and topics and improving relevance. Let's start from the beginning.

1. Your pages' focus: keywords vs. topics

The very first question you should be asking yourself when you think of embracing semantic SEO is this: How do I build my content? Should I (a) create pages around individual keywords, or (b) focus on broad topics and cover them in-depth? From the SEO perspective, these are the two (rather marginal) approaches to creating content today: (a) is the old-school way that you're probably used to, and (b) is the new-school approach that's becoming increasingly popular with the rise of semantic search.

If you took the old-school approach, you'd come up with tens of similar pages: a separate page (even if it's just a few sentences long) for each of the queries above. The problem with this is that in 2017, this kind of content will hardly ever be considered comprehensive, or even remotely useful, particularly in competitive niches. More likely than not, you'll be outranked by competitors who offer more comprehensive answers.

The new-school, topic-based approach implies creating a single page that covers all these topics, aka "The ultimate guide to buying vinyl". The basic idea behind the new-school approach is that you shouldn't be worrying about keywords at all — instead, you should build a comprehensive, original, high-quality resource, and Google will figure out the rest. Alas, for the time being, it's unlikely that it will.

You see, your "ultimate guide" may rank for the more generic terms like "how to buy vinyl". This is the kind of a broad term where the searcher isn't researching a specific aspect of the process, but rather researching the entire process and looking for the steps. Even if you include paragraphs on "where to buy rare records", Google won't always be able to figure out that that's something you should also rank for — simply because you're focusing on too many things with one page, and eventually each concept's prominence (or each keyword's term frequency, if you will) is diminished due to the length of your content and the number of micro-topics you're trying to cover.

2. Relevance

Now that we've figured out you need keywords to understand searcher intent and create content that matches it, it's time to move on to the role of keyword research and targeting in semantic SEO.

The first thing Google does when it receives a query is go through its index to find the pages that match it, likely using the TF-IDF algorithm. The process isn't always straightforward: the query may have to go through several refinements before Google retrieves possible search results from the index, and these results may be then further filtered according to various relevance and quality signals… And while it's true that in 2017, you can rank in Google for a keyword that's not even mentioned on your page, it only makes sense if you like to have things the unreasonably hard way.

Using keywords in your content helps inform search engines that your page is related to that term; in other words, it significantly boosts your chances of becoming one of the search results that will be considered for being ranked for the query.

In the age of semantic search, keyword research may have gotten less straightforward, but no less important. The two underutilized sources of keyword ideas that I feel give the best suggestions, particularly in the context of semantic search, are Google Autocomplete and Google Related Searches.

3. Meta-Relevance, Latent Semantic Indexing, and RankBrain

By now, Google's got a bunch of pages that it initially selected as potential matches to the query (with relevance 1.0). But how does it determine which results better fit the searcher's need and are more relevant to the intent behind the keywords? That's where semantics comes in.

LSI, or Latent Semantic Indexing, is a technology Google uses to understand the relationships between words, concepts, and web pages. By analyzing billions of web pages and the terms used in them, Google learns which terms are related, which ones are synonyms, and which commonly appear in the same context. This, in turn, lets the search engine build expectations as to the terms that are likely to appear in a given context.

So in a sense, both RankBrain and LSI are geared towards figuring out whether a page covers the topic thoroughly. But does thoroughness translate into rankings? Backlinko did a massive study to measure this. In it, they used MarketMuse to examine 1 million (!) Google results and the correlation of their topical authority (i.e. thoroughness and depth of expertise) and rankings.

4. Becoming a Knowledge Graph entity.

Google's semantic search is powered by the Knowledge Graph in numerous ways. The Knowledge Graph is a collection of entities - specific objects that Google knows a few things about, such as persons, places, and things. The Knowledge Graph's impact on search results stretches far beyond the branded panels that are sometimes displayed to the right of organic listings. Knowledge Graph data is used in organic rankings, rich answers, and various query-specific types of search results. One such type that seems to be gaining momentum is the "carousel" displayed for queries that name a certain category that a bunch of entities belong to:

If you do find an entity but aren't completely happy with what you see, go to Wikidata and use the search bar to find the listing about your company. Here, you'll be able to edit the details about your business, such as its description, official website, etc.

5. User Experience

The role of user signals in SEO is controversial, and this article isn't the place to debate it. In the context of semantic search though, it's crucial to understand that the fairest measure of the effectiveness of any new component in Google's ranking algo (be it RankBrain, Hummingbird, LSI, or anything else) is user satisfaction. Satisfaction may be measured with metrics like SERP click rate, time on page, and bounce rates. There are two ways Google obtains these metrics: through Search Quality Rating and real-life experiments in Google search. The scale of the latter is surprisingly vast; Google's Paul Haahr mentioned that whenever you run a Google search, you're in at least one experiment.

Now, if you think about those metrics and the factors they depend on… Sure, it's the quality and comprehensiveness of your content, which we've already discussed. But the list goes way beyond that: page speed, usability, readability, and mobile friendliness are just as important. Let's see how you can make sure your pages deliver the best user experience.