How many times have you searched for something in Google, and it seems to know what you want even before you have finished typing your search? It even seems to understand when we miss, what may otherwise be, critical information from the question. All of us will be familiar with this, and it is due to something known as semantic search.

What is semantic search?

As a digital marketing agency in Thailand, we are often being asked this question. Essentially, semantic search is a method of retrieving relevant information used by the leading search engines. In the past search engines had focused on keywords and their SERPs related to these keywords. Now companies such as Google, Yahoo and Bing try to understand the meaning behind the query.

The term semantics is one that all linguists will be aware of and means “to study meaning”.

Why is semantic search critical for producing relevant SERPs?

Semantic search is complex, and it needs to contend without numerous variables but why it is required, is quite straightforward. It is common for users to use different terms when they search for the same thing. In fact, it is equally as common that they use the wrong terms or don’t know how to articulate a query in the first place. The user knows what they want or mean, but they are having difficulty expressing it to others, including search engines.

One example that is frequently used is when you hear a song. You don’t know its title or the artist, but you did pick up on some random lyrics – which once again may not be 100% correct. You type the lyrics into your search engine and variations thereof, and the majority of the time you will find the song that you are looking for.

The complexity of the query is heightened when you ask the same question to Siri, Google Assistant or Alexa. They now need to understand your question or, in some cases, conversation, and the context of the problem. From here, they must compare it with the content that they have stored in their index. Quite a challenge we are sure you would agree, but this is only the beginning when it comes to its potential.

Searches can be polysemous

Very few people put intentionally ambiguous searches into Google. The biggest problem faced by search engines is that 40% of words used in the English language are polysemous meaning that the same word has more than one meaning. An example is the word “python” which has over half a million searches in the US each month. If we searched for that at Phoenix Media, we would probably be referring to the programming language, whereas the man in the street may be referring to the snake and man you met in the bar last night may be referring to Monty Python.

All of these are, of course, valid uses of the word “python” but it emphasises the importance of understanding the context of the query. It is something that is complicated further by the fact many nouns can also be verbs and/or adjectives. Throw in sarcasm and inferred meanings, and the query could be difficult for even a native English speaking human to understand, let alone Google or another form of AI. Semantics won’t work without an understanding of the context.

 Lexical hierarchy and entity relationships

Lexical hierarchy relates to the relationship between words with the word “partner” being a prime example. Partner is the lexical hierarchy superordinate to wife, boyfriend or spouse and so on. Entities could be a person, place or a thing and often has a specific role or job and can relate to the people associated with them. All very complicated and involves many language intricacies.

A commonly searched theme is for “partners” of characters in films and movies, for instance, Batman or characters in Star Wars. For Google to give you the answer that you want (remember, you may be referring to a different partner), it must understand all the terms. For instance, does the word “partner” refer to a romantic partner or partner as in Robin with Batman? Also, these films are a series, so different partners may be involved and played by different people. There is vast room for misinterpretation at many levels.

Google recognises personal interests

We often read about Google and social media knowing too much about us as they collect personal information. While few would disagree that on occasions it goes too far, without it, semantic searches wouldn’t work. Referring back to the python example, Google is aware that we frequently search for programming matters and have an active interest in digital marketing. As such, it correctly gives accurate and personal search results relating to the programming as mentioned earlier.

Location is also something else that Google can detect. Again, providing localised results is incredibly useful. If you were in Bangkok and searched for “language school”, results for schools in Tokyo or New York would almost certainly be useless. Another example of Google being aware of trends would be with the current coronavirus pandemic. There have been many coronaviruses throughout history. However, when people search for this term now, Google assumes that they are referring to Covid-19 – the current strain.

Google’s technology and semantic searches

Most of us a familiar with Google’s algorithm updates which are designed to improve the search engine’s understanding of language and context. At present, there have been four significant technological changes that play a vital role in semantic searches in 2020. These are:

  • Knowledge Graph– this was released in 2012 and was the birth of semantic searches. Knowledge Graph relies on two forms of information, structured data and the entity which is capable of extracting from the text
  • Hummingbird – this was an algorithm update in 2013 which saw the first shift away from individual keywords towards entire queries and recognising topics rather than merely words
  • RankBrain– this is one of Google’s most sophisticated systems and understands the meaning of words in context. This is a simplistic definition, but it effectively learns algorithms to satisfy search intent
  • BERT– BERT was introduced in late 2019 and handles around 10% of queries, generally the most complex. It specialises unravelling ambiguity and nuances and is one of the most advanced systems available when it comes to understanding context.

Semantic Search & SEO

As with many aspects of SEO, the rules are continually changing and evolving and what was applicable two years ago is unlikely to be in vogue today. Here are some of the key factors that you need to consider regarding your SEO.

  1. Focus on the topic rather than individual keywords

In the past, SEO companies in Bangkok would recommend writing similar pieces of content targeting slightly different keywords, and the results were impressive. However, in the modern era, Google recognises that many longtail keywords have effectively the same meaning, so groups them together. Therefore, it is recommended that a topic is covered in depth rather than just one aspect, or more accurately, keyword. Articles that rank highly may well include hundreds of keywords, but they will be relevant to the topic.

However, it is advisable to remember your audience and not go into so much detail that the reader loses interest which brings us perfectly onto the second point.

  1. Consider search intent

Remember that for most users, they only want enough detail that will enable them to complete their task. For instance, if they wanted to know how to change a tap, they are not interested in hot water systems or anything else which may have a loose connection to changing a tap. In fact, if there is too much detail, you are likely to have a high bounce rate as the user searches for a more relevant answer.

  1. Semantic HTML

When the internet was first conceived, it was a series of interlinked documents that had little meaning. Nowadays, as we hope we have highlighted, context is essential, and that starts with the HTML. HTML adds sense to code while HTML5 provides semantic elements and most CMS themes use this, or you could use a plugin. Sadly, it only works to a certain degree, and this is where schema markup comes into play.

  1. Schema markup

Schema marker is often referred to as structured data, something that we touched upon earlier. Schema markup is a method of marking pages and is the most common semantic framework available for the web. has various types and associated properties that can be used to mark your content and make it quickly detected, extracted and understood by Google’s algorithms.

  1. Brand building

Your objective is to build your brand in such a way that it becomes a Knowledge Graph entity. It is arguably the hardest element of the semantic search SEO to turn into reality, and it is the result of brand building and organic SEO. It is an area which should you require more information about; our team will be able to explain in terms relative to your company and website.

  1. Use links to build relevance

Finally, one thing that hasn’t changed over the years is the fact that links are still an accurate indicator of relevance. All your content should have internal and external links as well as natural anchor text. Google will quickly be able to understand the text and put it into that all-important context, meaning it will rank higher on SERPs.

More information

If you would like more information on semantic search and how it impacts your SEO, get in touch with us today.

  • Published on : Sunday June 14, 2020

About the author

As the managing partner for Phoenix Media, Rob brings over 10 years’ experience in digital marketing and running successful agencies in the UK, Australia and Thailand. Starting in a sales role he has covered all aspects of the agency from sales and service to technical ad operations. Reach him directly on

Recent blog posts and insights.