Wednesday, 14 May 2014

How to Get Search Traffic from Google’s Knowledge Graph

When Google rolled out the Knowledge Graph on May 16, 2012, it caused quite a stir in the search community. According to the reports at the time, Google took a step into artificial intelligence and got a thousand times smarter.
Even though the Knowledge Graph has been out for two years, there are still very few SEOs who really grasp the significance of Knowledge Graph Optimization (KGO).

The very fact that Google is ‘smarter’ means that you need to be smarter too. I don’t mean ‘smarter’ in order to game the system, but ‘smarter’ in order to better understand how users can get the things they want when they search.
Google’s goal is to give users the most accurate search results in the shortest amount of time. Your goal as a marketer is the same — to give users precisely what they are searching for. Obviously, you want those results to be your own, which is why you need to be optimizing your web presence for the Knowledge Graph.
But before we get started, let’s take a look at what the Knowledge Graph is:

What is the Knowledge Graph?

The Knowledge Graph is Google’s way of answering questions and delivering information in a direct way.
The best way to understand it is to search Google for “what is the knowledge graph?” The result you will see is a product of the Knowledge Graph, presented at the top of the SERP in larger type and framed with a gray box.
knowledge graph
To be more specific, the Knowledge Graph (KG) - according to Search Engine Land - “understands facts about people, places, and things and how these entities are all connected.”
Seeing KG in action helps understand how it works. Below are some examples.
Searching for “famous actors” turns up this Google SERP. The picture carousel at the top is a Knowledge Graph result:
knowledge graph actors
In the search for “Colorado Rockies” below, the SERP is virtually dominated by Knowledge Graph results:
Searching for “best sandwich shop in San Francisco” shows another carousel result plus a map result, both products of KG.
Results that have to do with individual people will vary according to how you word your query. For example, if I type in “who is the ceo of google,” this is what I get:
But when I search for “who is Larry Page,” this is what I get:
larry page
The Knowledge Graph provides personalized results based on location and other factors as well as results from your Google+ account. Here is an example of the personalized results that it will fetch:
ip address

How the Knowledge Graph works

Under the hood of the Knowledge Graph, you’ll find several fascinating algorithmic features. Here are a few of them:
  • Semantic search – the driving force behind Knowledge Graph’s input understanding is semantic search. KG considers a variety of data points — word variation, synonyms, concept matching, natural language, IP location, and search context — to deliver more specific results. Many of Google’s recent algorithm updates have further pushed the power of semantic search.
  • Entity indexation and disambiguation – the Knowledge Graph is a grand attempt to catalog every thing — i.e., noun, object, or entity — and connect it to every other thing. These “things” are technically called “entities.” The process of entity mapping is called Entity Recognition and Disambiguation, or ERD. As we’ll see below, understanding the power of entities is a key component of KGO.
  • User behavior – what I find interesting is that the Knowledge Graph depends heavily on user behavior. In response to the query above “what is the knowledge graph,” the graphed SERP comes from Wikipedia, not Google, even though “knowledge graph” is a branded term of Google. This fact alone shows that Google is algorithmically preferring (as far as we can tell) Wikipedia results over its own page, in part, due to the fact that Wikipedia is the go-to resource for most users. Search traffic, including CTR, organic, and direct, indicates that Wikipedia contains the best answer to a user’s query.
Now that you understand what the Knowledge Graph is and how it works, let’s break down what you need to do in order to get more search traffic.

Continue your content marketing using entity keywords

At the most basic level, use keywords. The Knowledge Graph, as an entity recognition and disambiguation system, thinks of your keywords as “entities.” It attempts to recognize a particular entity, disambiguate it, and then relate it to the millions of other entities within the database.
Without the use of keywords, it would be very difficult for Google to connect pages on your website to specific search queries.

Use schema markup everywhere possible

Schema markup is the quickest and most effective way to help the Knowledge Graph return your results in graph boxes. We already know that it shows in 36% of all Google search results. You can infer that Google prefers pages that have schema markup. Thus, you should be using it wherever you can.
Markup options are available for hundreds of entities on your website. Since the Knowledge Graph depends heavily on this markup to produce helpful results, you should be using it for your site entities.
You’re probably familiar with markup for, say, movies:
However, there are many other schemas that you can use on your site.
Have a recipe? You can use a markup schema to identify the method of cooking, the type of cuisine, the length of time it takes to prepare the dish, and the ingredients that should be used.
Google may not feature Knowledge Graph data for every schema data point you input, but it will return some. And, as semantic search and markup become more important for the algorithm, it will probably return more as time goes on.
Notice how, in the results below, markup is providing Knowledge Graph results. Fundamentally, Google reads these as recipes, even in the absence of markup. But beyond that, it provides KG results for the markup identifying prep time, calories, ratings, and reviews.
stir fry
This is just one example of the type of markup that produces Knowledge Graph results. There are hundreds of other schema options that you can use. You can find a variety of schemas at

Optimize your Google+ pages

The Knowledge Graph draws much of its information directly from Google+. Businesses and organizations with optimized Google+ pages have an advantage for KGO. To optimize for Google+ is to optimize for the Knowledge Graph.
Here’s an example. If you are looking for a good place to eat in San Jose, California, you might search for “eatery in san jose.” The Knowledge Graph carousel populates photos, reviews, and directions directly from the Google+ pages of the relevant food establishments.
When you click on a specific business, e.g., Flames Eatery, the query input changes automatically (notice the search field), and the SERP is now populated with even more Google+ data. If you are logged in to Google while searching, you can “follow” this restaurant by adding it to your Google circles, or you can write a review.
Google+ allows you to set up your business’ hours of operation, menu, and other important details. Searchers will see this in the Knowledge Graph results.

Invite customer reviews

One of the most common KG results is the reviews. Google depends on the review text to create graph displays in search and map search results. See how user input affects Knowledge Graph data in Google maps:
ruby skye
Clearly, detailed reviews with plenty of information provide the most helpful results.
The art and science of inviting business reviews is outside the scope of this article, but keep in mind that they do impact search relevancy and, more to the point, Knowledge Graph results.

Use Freebase MIDs (sameAs)

Freebase is an entity database that assigns an MID (Machine Identification) to every one of its millions of entities. Anything that is relatively well known probably has an ID in Freebase.
Google, with its purchase of Metaweb in 2010, acquired the Freebase database. Although Freebase remains a Creative Commons licensed organization, the data it has produced is subsumed entirely under Google’s Knowledge Graph database. Thus, connecting your brand entity with Freebase provides a way to correlate structured data with Freebase data, aiding entity disambiguation for streamlined KG results.
The sameAs property is one of the ways that this verification works. AJ Kohns describes it as an “entity canonical.” Adding the itemprop=”sameAs” markup will instantly provide the disambiguation that is needed to deliver a straightforward Knowledge Graph display.
If your entity (person, thing, business, etc.) is not part of the Freebase database, you can easily add it.

Help Wikipedia

One of the most obvious sources of data for the Knowledge Graph is Wikipedia. You can help contribute to Wikipedia’s accurate, reliable, and helpful information. Wikipedia, however, is not a means to SEO/KGO, but rather a data source for the Knowledge Graph, which is a component of a successful web presence.
Like I mentioned above, KGO is not the same as search engine optimization. It requires a more detached and user-information motivated view of online marketing.

Help Google

Finally, you can actually help Google refine the results of the Knowledge Graph by labeling them as “not useful,” “useful,” or “awesome.” This is probably the least effective way to influence the results, but I wanted to at least mention it in closing.
Here’s how you can tell Google what you think.
Within a Knowledge Graph box, usually at the bottom, you can find the word “Feedback.”  You can click on the word and tell Google what you think about the results.
Occasionally, you will receive a pop-up asking you for your response:
Personally, I’m skeptical of how much of an impact my responses matter in the grand scheme of all things Google. But, hey, if I can help Google improve its search capacity, why not?


Knowledge Graph Optimization is less about quick wins and slick tricks and more about high-level trustworthy marketing efforts. The core of KGO is giving helpful information to users and letting them make the best decision.
How else can you optimize for Knowledge Graph results?

No comments: