Sometimes it’s hard to see just how smart technology is becoming. The advent and evolution of Artificial Intelligence — already more than 60 years old, by the way — is something that we insist on thinking about in futuristic terms. The reality is that AI is here, and you interact with it every day.
One way that most of us interact regularly with AI is through Google Search. You might think that your search query is simply parsed into keywords so that google can return popular web pages that contain those keywords, but that’s a 20-year-old concept of search.
Since the Hummingbird algorithm update in 2013, the name of the game is semantics - and Google Search is getting scary good at it.
What is semantic search?
If someone were to ask you to recommend a restaurant, your answer would most certainly be affected by the context of the question:
- What city are they in?
- Are they looking for something formal or casual?
- Does the person asking have dietary preferences or needs?
- Is it karaoke night somewhere?
Generally, the more context you have, the more accurate and helpful your answer can be. This is semantic search in a nutshell. Your search query is being processed in light of the context of the question.
Here are some examples of context that Google now considers alongside your search:
- The Query: Keywords are helpful, but Google also considers the query as a whole (the way you ask vs. what you ask).
- You: Google looks at your search histories, emails (ahem, Gmail users), social activity, spending habits, political leanings, etc. Google knows more about you than you think.
- Your Surroundings: Google uses your location, time of day, device information, and recent activity to guess at your intent.
- Natural Language Processing: Google is trained to find connections between topics and queries by listening to how we naturally speak.
To see semantic search in action, all you need to do is search for “restaurants near me”. You’ll notice that the results you get will be influenced by your location, time of day, and maybe even the fact that you recently tweeted about craving lasagna.
How does all of this affect search engine marketing?
That’s the million-dollar question, isn’t it? What strategies can we employ to work best with an increasingly human-like search algorithm?
1. You need to make it easy for Google to accurately understand your content.
Structured data is an evolving strategy to help search crawlers understand and process our web content. Structured data is hidden code that we can use on our web pages to tell crawlers what our content is about. [Recommended Reading: What is Structured Data, and Why Should I Care About It?]
Some estimates put Google’s data footprint in the neighborhood of 15 exabytes (that’s 15 billion gigabytes). Obviously Google doesn’t comb through all of that data for every single search (40,000 per second, incidentally).
Google uses its crawlers to read your web pages and organize them for fast responses to common queries. We can’t control how Google decides where we get filed, but we can give Google as much helpful information as possible and increase our chances of being filed accurately in a position of authority.
2. You need to tightly map your content together into topic clusters.
Content mapping and internal linking are proving to be important factors as search gets more semantic. This goes back to our first point about being filed by search crawlers.
The goal is to break your website down into topics and group pages together based on those topics. In addition, your internal linking should support this topic architecture. The end result should be a body of content that is well organized by subject matter — not publish date (as is common with basic blogging strategies).
This content mapping strategy allows for search crawlers to navigate your site in a way that suggests contextual authority (i.e. moving naturally from page to page within a single topic).
3. You need to incorporate semantics into your content writing.
There’s a particular exercise that we inflict on ourselves every time we put together a content mapping strategy. Once we have an understanding of our topics and how they are mapped together, we list as many different semantic search queries for each topic as we can think of.
Each semantic query combines a specific buyer persona with a specific search scenario. Obviously this can result in a significant list of queries for a single topic, so we narrow it down to the top 5-15.
This exercise is tedious, but it essentially serves as the foundation for our content strategy for months or even years. Our approach to content is that a single page should serve a single semantic query as effectively as possible.
By using semantics to inform our writing and planning, we can create content that is perfectly suited to a small number of our ideal users — rather than imperfectly suited to a large number of less-than-ideal users.
If semantic search means Google responds to our queries in a more human, contextual way, then it seems reasonable to adapt our SEM strategies to be more human as well.
The now-dated approach to content that says “more content = better results” worked when it was the early adopters who were using it. But now that we’re drowning in “more content” from the late majority, we see that what we really need is more context for the content we already have.
Thus, the human approach to SEM would be to focus on context over content by revisiting and re-optimizing the content we already have to be:
- Easier to crawl
- Accurately mapped in topic clusters
- Internally well-linked, and
- Written with a focus on individual personas and scenarios.