Just over a decade ago, you could visit Google.com and type in “earth mass in tones” (yes, typo included) and get this result:
That’s right: the #1 result is a ringtone website (who remembers those?) because Google noticed the word “tones” in your query (and that result likely had “tone” blasted all over its content and backlinks).
But you really meant to type in “tons.”
As in “earth mass in tons.” Too bad Google had no way of understanding the meaning of your words back then.
Today, this is what the query “earth mass in tones” (typo included) looks like:
Not only does Google account for the typo, but it produces the answer directly within search results, along with additional information related to the earth’s mass.
How? Semantic search.
Today, search engines like Google, Yahoo, Bing, and Yandex think and learn like real humans. They can process natural language, understand the relationships between words, and produce an answer to your query even if the keywords you use don’t exist anywhere on a page.
And there’s only one way to succeed in semantic search: semantic SEO.
In this article, we’re going to explore the underpinnings of semantic search, how Google and other search engines evolved to understand meaning, and what you can do to optimize your SEO strategy for the semantic web.
Subscribe to our weekly newsletter for tips so good that we might put ourselves out of business.
What is semantics?
To understand semantic SEO, you need to understand semantic search. But to understand semantic search, you need to understand the basics of semantics first.
Semantics refers to the study of language.
More specifically, the study of the subtle shades of meaning within language.
Semantics analyzes context, sentence structure, word hierarchies, and relationships between words to identify the intended meaning of a word, phrase, sentence or text. It explores the nuance of language to arrive at the truest meaning of what’s being said.
I know—riveting, right?
But semantics is the cornerstone of semantic search, and, by extension, semantic SEO.
What is semantic search?
Semantic search follows the same principles as semantics, only executed by search engines, not humans.
Semantic search refers to a search engine's ability to analyze and interpret search query nuance (from context to search intent to relationships between words) so they can produce the most relevant and accurate search result possible.
Why is semantic search important?
Semantic search marked an evolutionary leap forward for search engines.
Because the words and phrases people use to search for information online are often unconventional and ambiguous, and natural language is loaded with complexity. Not to mention, the most relevant search results need to reflect personal preference too.
Before semantic search, search engines had no way of interpreting the growing complexity of search queries. Results suffered. It usually took 3-4 searches to find what you were looking to find.
Take the keyword, “what’s the name of the bad guy in Dune?”
This query seeks a specific answer (Vladimir Harkonnen), but since the searcher doesn’t know the name of the villain, they replace his name with “bad guy.” Close enough, right?
Prior to semantic search, algorithms didn't have the capabilities to interpret that query. It would have taken you a handful of searches to find the right answer, maybe more.
Today, because of semantic search technology, Google not only knows who you’re talking about when you say “bad guy in Dune,” but they provide the answer directly within search.
Pretty remarkable, right?
Unlike traditional search of the past, semantic search is capable of providing the answers you intended on finding, regardless of what keywords you typed in Google.
History of semantic search
To fully understand how search engines evolved from a web of strings (keywords) to a web of things (connected topics and understanding meaning), it helps to look at the technological developments that led the way.
More specifically, Google’s algorithm updates.
The 2000s (pre-semantic search)
Up until 2011, search engines like Google relied heavily on keywords to rank pages. Like, literal strings of disconnected words, regardless of the meaning or intent those strings of keywords combined to form.
For example, if you wanted to rank for “drumstick sizes,” you would need to use “drumstick sizes” in various places on a page, whether or not you actually provided a thorough list of drumstick sizes. Even still, there was a good chance you’d also rank for Neslté’s drumstick ice cream.
That’s because Google had a hard time understanding the intent of your search; they just processed keywords as their own, unconnected strings of characters. To them, “drumstick sizes” didn’t mean “I’m looking for a chart that describes the different drumstick sizes available so I can choose the best size for my hands.”
This created two problems:
- Search rankings were easy to game. Just stuff keyword onto a page and voilà: page one rankings.
- Search results were often irrelevant. Search engines couldn’t identify context or intent, so if you typed in “Rio” (and were thinking about the city in Brazil), they may produce results for “Rio” the casino, or “Rio” the movie. See the problem?
Knowledge panel (2012)
The challenge: Google really needed to understand the world more like people did, not like computers.
Algorithms needed to know that a search query wasn't just a string of random words, but that those words were talking about real things with real meaning and distinction from other things.
In 2012, for the first time, Google started collecting data about “entities” (persons, places, things, or objects) from open source databases (e.g. Freebase.com, Wikipedia.com, CIA World Factbook), licensed databases, and special coding called “structured data” (e.g. Schema.org), then organizing that data into their own structured database called the Knowledge Graph.
Google’s intention was to build an organized database of information and facts that are verified and definitive. Then, using machine learning (technology that learns on its own based on data inputs), it could use that database to learn how different concepts and entities were connected, and, as a result, better understand the meaning of your queries.
Today, Google’s Knowledge Graph has amassed over 500 billion facts about 5 billion entities.
For queries that they’ve collected definitive information about, they use Knowledge Graph data to influence search results. Most notably in the form of a Knowledge Panel.
Hummingbird update (2013)
The Knowledge Graph took the first semantic step forward.
But with the rise of mobile search, conversational search (AKA voice search), and natural language queries, and thus more complex queries, Google needed a better way to understand search intent (i.e. what people actually mean, regardless of how they say it).
So Google applied the same machine-learning technology they applied to Knowledge Graph data, only to the billions of pages across the internet, with the expressed intent of understanding the meaning and context of complex queries.
They called this core algorithm update “Hummingbird” because of its ability to provide “faster and more precise” answers.
Think of Hummingbird like a total replacement of Google’s search engine, built for semantic search: smarter, faster, and more capable of discerning the true meaning of words.
Most notably, Hummingbird marked a monumental shift away from keywords and keyword density (strings) toward topics and search intent (things).
It’s not that keywords didn’t matter post-Hummingbird; it’s that unnatural placement of keywords for purposes of ranking has become irrelevant. Now, it’s more beneficial to write like a human and to provide the best answer to a query.
In 2016, Google took semantic search one step further.
Whereas Hummingbird was an overhaul of the core algorithm that powers Google (like a brand new engine), RankBrain was an advancement that made it better (like a new part in the engine).
Initially, RankBrain only processed 15% of queries. Today, all queries pass through RankBrain.
Like Hummingbird, RankBrain uses machine-learning and AI to better understand the semantics of a search query and provide the most relevant results. It’s another step forward toward analyzing the topics, subject matter and search intent behind a query, not just literal strings of words and phrases.
RankBrain incorporates things like historical search data, location data, content freshness, and other personalized data not connected to a landing page to understand intent and serve hyper relevant results.
Remember our drumstick example?
With RankBrain, now it’s easy for Google to produce an amazing result when you search for “drumstick sizes” even if no page includes that exact phrase.
Not a single instance of “drumstick sizes” appears on this page. It doesn’t need to.
Thanks to Hummingbird and RankBrain, Google knows that when someone searches for “drumstick sizes,” they’re looking for information on how to choose the right drumstick, regardless of what keywords they types in the search box.
Google’s latest advancement in semantic search comes in the form of BERT (Bidirectional Encoder Representations from Transformers). BERT is an algorithm update that analyzes search queries, not web pages.
Cracking your queries
So that’s a lot of technical details, but what does it all mean for you? Well, by applying BERT models to both ranking and featured snippets in Search, we’re able to do a much better job helping you find useful information. In fact, when it comes to ranking results, BERT will help Search better understand one in 10 searches in the U.S. in English, and we’ll bring this to more languages and locales over time.
Particularly for longer, more conversational queries, or searches where prepositions like “for” and “to” matter a lot to the meaning, Search will be able to understand the context of the words in your query. You can search in a way that feels natural for you.
BERT builds upon the machine-learning of Hummingbird and RankBrain, this time targeting the 15% of queries that Google can’t anticipate because they’ve never been searched before.
BERT takes natural language processing to the next level: at its core, BERT helps Google better understand the meaning of each word within a short or long phrase and how those words relate to each other within the sentence.
Same ol’ semantic song and dance, only improved.
Here’s an example of an improved search result post-BERT:
Semantic search didn’t happen overnight.
It evolved over the last decade, thanks to advancements in machine-learning, natural language processing, and artificial intelligence.
Although the path was long and windy, something tells me we haven’t seen the last evolutionary step in semantic search.
Examples of Semantic Search
Have you noticed that the appearance of SERPs (search engine results pages) has changed quite a bit over the last decade?
That’s because semantic search entails more than just 10 organic results. In fact, Google’s ability to understand the meaning and intent of queries makes it possible for them to serve additional information within SERPs in loads of different ways:
- Featured snippets
- Content blocks
- Related searches
- Autocomplete suggestions
- People also ask section
People also ask
What is semantic SEO?
Ok, so if semantic search refers to a search engine’s ability to think more like humans and understand the true meaning of what someone is searching for (regardless of how they ask), then produce myriad results based on that understanding, what do you think semantic SEO refers to?
Semantic SEO, by extension, refers to the process of making your content more human and understandable by focusing on topics and satisfying search intent, not by obsessing over keywords and phrases.
Why is semantic SEO important?
Semantic SEO marks a fundamental shift in the way SEO specialists do their job.
Today, semantic SEO is about optimizing for topics, not keywords. Or, as Google would say it, optimizing for entities and their related attributes, not keywords.
For example, if you want to rank for “Mona Lisa,” in order for Google to know with certainty what you’re page is about and how relevant it is to the query, they would expect the page to also include information about Da Vinci (who painted it), Louvre museum (where you can see it), Paris (the city it’s located in), and La Joconde (“Mona Lisa” in French).
Let’s use this article as another example.
The entity I want to rank for is “semantic SEO.”
In order for Google to know with a high degree of confidence that this article is about semantic SEO, it’s not enough just to provide a simple definition or include the words “semantic SEO” over and over again. Not anymore.
Instead, I need to look at the related attributes and subtopics associated with semantic SEO, then incorporate those attributes and subtopics within the article. Which, in this case, would mean talking about semantic search, Hummingbird, BERT, RankBrain, search intent, and the Knowledge Graph.
See the difference? Optimizing for topics and their subtopics is a massive departure from optimizing for keywords.
How to optimize your semantic SEO
Semantic SEO isn’t difficult.
In a lot of ways, semantic search just rewards common sense and quality.
But here are seven non-negotiables for better semantic SEO:
- Identify and satisfy search intent
- Explore topics, not keywords
- Build your site like Wikipedia
- Don’t obsess over rankings
- Use structured data
- Use semantic HTML
- Listen to Google SERPs
1. Identify and satisfy search intent
Plain and simple: ranking at the top of semantic search results comes down to understanding the meaning behind the words people use to search. Not necessarily what they say, or how they say it, but what they mean.
Forget what keywords they literally type in; what do those keywords actually ask? What information do they intend to find based on that query? What questions do they imply, even if not explicitly stated? What format do they expect: a list, tool, long-form article, something else?
Intent rules everything in semantic search. Satisfy it.
2. Explore topics, not keywords
Like we mentioned earlier, it’s not that keywords don’t matter anymore. They do.
It’s that focusing solely on a phrase or string of keywords isn’t going to help you rank higher in semantic search.
Remember entities? Since moving from strings to things, Google is looking for entities and their attributes. Which means thorough answers, regardless of the keywords used on the page.
Think more broadly about a keyword: what is the topic (entity) your keyword seeks to explore, and what other semantically-related subtopics (attributes) would you expect to find in that same family? Then go deep.
Use keywords as a starting point. But articles should incorporate a family of keywords, not just one:
- Seed keywords (short, broad, high search volume, competitive)
- Long-tail keywords (long, specific, lower search volume, less competitive)
- Semantic keywords (words you would expect to find associated with primary topic)
- Keyword variations and synonyms (e.g. “seed keywords,” “head keywords,” “short-tail keywords” all refer to the same thing; keep them on the same page)
This accomplishes two goals. First, adding depth creates a better user experience for visitors. Second, focusing on topics (not keywords) forces you to explore more attributes of an entity, making it easier for Google to identify the meaning of your page.
3. Build your site like Wikipedia (aka hubs and spokes)
Build your site like Wikipedia: hubs and spokes.
Hubs and spokes? Yup. One primary topic (entity) that links to dozens more subtopics (attributes).
Think about it: Google uses Wikipedia.com structured data as a primary source for their Knowledge Graph. Which means they trust the website’s information and architecture so much that they’re willing to use it as a definitive and verified resource center, plug it directly into their Knowledge Graph, then consume that information to learn how different topics (entities) and subtopics (attributes) are connected.
That’s because Wikipedia is literally built around entities and attributes already, and intricately woven together with internal links (links from one page to another on the same domain).
If Wikipedia’s site structure is good enough to help Google learn the meaning of topics and their relationship to other topics (and fuel semantic search), don’t you think your site should do the same?
Each Wikipedia page is like its own entity (topic), and each entity internally links to hundreds more relevant attributes (subtopics), which in turn function like their own entities with attributes.
By building hubs and spokes (like a Russian nesting doll), Wikipedia is able to create a big web of meaning for visitors and search engines to easily navigate and understand.
Be like Wikipedia. Disambiguate your content by linking deeper to more relevant content.
4. Don’t obsess over keyword rankings
Keyword rankings. Meh.
Keywords still matter (this is SEO), like we mentioned twice now.
But do keyword rankings matter in semantic search?
Kinda? Not like they used to, that’s for sure.
Think about it: If Google can serve the best answer to a search query (AKA keyword), even if those keywords don’t appear anywhere on the page, they’re essentially rewriting search queries. And if they rewrite search queries, how will you know which keywords to track anyways? You won’t.
Sticking with our previous example, if you wanted to track your rankings for the keyword “Vladimir Harkonnen,” but Google also showed your results when people type in the keyword “what’s the name of the bad guy in Dune?” (essentially rewriting that query to “Vladimir Harkonnen”), you would never know.
Tracking rankings for “Vladimir Harkonnen” wouldn’t tell you. So don’t obsess over them anymore.
Better to focus on traffic at the category level. If you filter and group URLs that explore different parts of the same topic, are you growing overall traffic, month over month, for that topic?
5. Use structured data (schema markup)
Structure. Structure. Structure.
The more explicit information you can provide Google about the meaning of your content, the better you’ll fare in semantic search.
Structured data isn’t a direct ranking factor. But speaking the language of search increases your chances of indirectly ranking.
How? Schema markup.
Schema.org is a semantic markup language founded by all the major search engines. With Schema, you can wrap certain elements on a page with HTML tags that provide search engines with additional information about the content. Visitors don’t see Schema, only web crawlers.
For example, you can use Schema to tell search engines that your content is talking about details of a car, information about health conditions, star reviews, recipes, people and their children, events, product categories, and so much more.
Before Schema microdata:
After Schema microdata:
See the difference?
Bonus: Google uses Schema markup like reviews, recipe details, and product information to create rich snippets (results with additional information next to them) in SERPs.
- Schema Markup: The Language of Search
6. Use semantic HTML
Semantic HTML is HTML that clearly defines the different sections of a page.
Like Schema markup, semantic HTML speaks the language of search engines and provides explicit information about the meaning of your content.
For example, popular non-semantic elements include <div> and <span>, both of which tell developers or web crawlers nothing about the content.
But semantic elements like <form>, <header>, <table> and <nav> explicitly tell developers and web crawlers what the different sections of your page are about.
7. Listen to Google SERPs
Google SERPs are loaded with semantic search features like the examples we mentioned earlier.
Don’t overlook them.
People also asked, related searches, autocomplete suggestions, featured snippets, content blocks… Google pulls this data into SERPs because it’s semantically-related to the query. They’re telling you that this information is important.
If you want to rank for a specific query, use these features to inform your content marketing strategy.
Answer the “people also ask” questions within your content. Use related searches as headings or subheadings. If there’s a featured snippet, ensure you provide a clear definition within your own content.
Wrapping up Semantic SEO
SEO has come a long way from traditional keywords (strings) to entities and meaning (things).
Gone are the days when you could search Google for a question about the mass of earth and get a ringtone website as the #1 answer.
We’re all better for it (well, maybe not the ringtone company).
For marketers to succeed today, you need to embrace semantic SEO:
- Obsess over search intent
- Explore broad topics, not keywords
- Build your own entities and attributes (hubs and spokes)
- Speak the language of Google (structured data)
- Don’t obsess over keyword rankings
- Listen to Google SERPs (they’re telling you something)
- Create high-quality content for people first, Google last