The world of search engines and search marketing is in a (rather large) state of flux right now. New algorithms like Hummingbird and RankBrain have taken machine learning, AI and natural language processing and applied it to search results. The result?
Something known as “semantic search.”
Semantic search is method by which search engines deliver more accurate and relevant search results based on the search intent and context of a user’s search query.
Google’s machine learning-powered RankBrain algorithm processes queries to determine what the user actually wishes to accomplish:
- To go somewhere. These users want to find a particular website or page. They just don’t know the URL.
- To learn something. These users want to answer a particular question (“is jon snow really a Stark”) or just general information on a particular topic (“jon snow”).
- To do something. These users are looking to complete a particular action when they click through from search results. Typically this means purchase something, but could also mean submit a lead gen form, sign up for a newsletter or download a file.
As mentioned above, the interpretation of query intent is handled by RankBrain. Google’s Hummingbird algorithm helps figure out which pages are most relevant to a topic and match best with the query intent.
Semantic search isn’t just limited to queries and landing pages, either. It’s also changed the way Google search results look. All those Knowledge Graph panels, Twitter cards, Answer Boxes and related questions?
Also part of semantic search.
What’s the Difference?
Well, for one thing, take a look at the old Google search results. Then look at them today:
Pretty big difference, huh?
But it’s not just cosmetic.
Think about what someone searching for “jon snow” wants.
It’s most likely to learn about his character biography, the actor who plays him or maybe some news about the show/fan theories about the character.
Under the old search engine system, you could wind up with a SERP full of results about buying Jon Snow costumes, Game of Thrones merchandise, or someone you can hire to play Jon Snow at your birthday party. However, the likelihood of any of those pages being useful to me, someone using the “jon snow” query, being useful to me is quite low.
That is what semantic search is all about: interpreting the intent behind the query.
How to Get Ready for Semantic Search
How do you optimize a website for semantic search? There are three main pillars on which to base your semantic SEO.
Since matching search intent with relevant pages is the main goal of semantic search, optimizing pages for that intent should be the main goal of your semantic SEO.
Start at the core of your business: your product. Use your product as a topic, and then think about the situations in which someone would be searching for it.
One of our products, for example, is an SEO audit that checks various technical aspects of a website that impact search engine optimization. One of those aspects is the XML sitemap.
So we start with topic of XML sitemaps and SEO, and then ask ourselves, what SEO need does someone searching for “xml sitemap” want to meet? You can answer this question through your own brainstorming, using a long tail keyword research tool (we think Answer the Public is a good one) or just do a quick Google search on your topic and see what appears in the related questions feature:
These are the questions my content should seek to answer (since I’m targeting informational searches at the moment).
Once you have your question to answer, there are some neat tricks to help Google understand how you’re helping your readers:
- Use the question as a keyword: put it in the page title and <h1> content. If you’re creating a how-to guide, put each step in the <h2> subheads, and try numbering them.
- Put the answer to each question in the body text directly after the <h1> tag. Try to avoid long intro paragraphs: Answer Box results are limited to 50 or 60 words.
Topical relevance is the second main pillar of semantic search. Which may not sound like much to you, since relevance has always been a part of search ranking. What’s different is how Google is measuring that relevance.
Thanks to developments in natural language processing and machine learning, simply using keywords a certain number of times simply won’t cut it.
But how can you measure that?
The best measure of relevance is to listen to your users — they’ll tell you with their on-site behavior.
Measure user interaction with your pages to see if it’s relevant to your topic and their queries by tracking these metrics:
- Time a user spends on your site
- Bounce rate (Take this with a grain of salt, though. Just because someone doesnt’ view another page doesn’t mean your content is irrelevant.)
- Event (clicking a button, watching a video, enlarging an infograph, etc.)
- Scroll rate
Use Your Knowledge Base
Bringing structured data, the technology that powers semantic search, onto your website will help Google understand what your content is about. Structured data refers to a way of storing information in a way that is machine readable.
Adding structured data adds context to a page
that helps search engines determine if it’s relevant to entities used in
a query. Entities are people, places and things stored as structured
Use structured data by marking up your entities use the Schema.org taxonomy and adding them to your pages using JSON-LD. You can use other methods of adding structured data, but these two are preferred (ever so slightly) by Google.