Disclaimer: When discussing patents, it’s important to remember that simply filing a patent does not mean a technology is in use or will ever be used. It is simply a strong indication that an idea is being considered and likely tested.
Every now and then, a patent comes across my radar that gets me excited, and one granted recently to Google fits that bill perfectly.
We’re heard repeatedly from Google that social interactions are not a search ranking signal. In fact, you can read a Tweet from Google’s Gary Illyes in response to the statement, “Some controversy over whether Google takes social into account for SEO….” His reply:
So the answer is “No,” right?
Maybe, and here’s where it gets interesting. Understanding that the folks at Google tend to give answers that are technically correct but not always in the spirit of the question asked, we can hear what’s being said as, “Social is not taken into account as a direct signal when determining the rankings.” This is very different than not being a signal at all, for reasons we’ll illustrate shortly.
Another aspect of Google search that we need to be constantly aware of is that RankBrain now applies to all queries. Essentially, this means that artificial intelligence (AI) is interpreting all queries to some degree. While at this time the AI implementation revolves more around using machine learning to understand the nature of the query (and likely type of content and format being sought), its rollout to all queries and the promotion of John Giannandrea to Head of Search at Google marks the push into AI control over larger portions of the Google algorithm.
“Why does this matter?” you might ask. The answer is that once machine learning and AI influence sections of the ranking algorithm and not just the query interpretation, there will forever be factors and functions and results that not even the engineers at Google can fully determine. The answers Google gives will no longer be based on an actually conclusive understanding of what’s going on, but rather a better than average guess — and the exact factors and their weights won’t be fully known.
I mention these things to wrap context around the patent we’ll now discuss. Google was recently granted patent US 2016/0246789 A1 titled, “Searching Content Of Prominent Users In Social Networks.” If it sounds incredibly interesting, that’s because it is.
Good! Let’s get started. To keep things clear, I’m going to go through some of the key points and sections of the patent. I’ll begin with what they’ve written and immediately follow that with my interpretation of what it means in real English, then discuss any implications for SEOs.
The abstract of a patent is essentially the outline of what it covers.
They wrote: Methods, systems, and apparatus, including computer programs encoded on computer storage medium, for receiving a query; retrieving one or more social restricts associated with the user, the one or more social restricts comprising a set of author-based query restricts; generating an augmented query based on the query and the set of author-based query restricts; obtaining a set of social search results that are responsive to the augmented query, each social search result in a first sub-set of the set of social search results being associated with an author-based document restrict that corresponds to an author-based query restrict in the set of author-based query restricts; and providing the social search results for display to the user.
What it means: Essentially, what they’re stating in the abstract is that this is a patent for the “methods, systems, and apparatus” for the following steps:
- Receive a query from a user.
- Retrieve for that user an adjusted set of results based on social connections and what those connections write, share, post, etc.
- Adjust the initial user query based on these social adjustments.
- Display a search result based in part on the initial query and adjusted based on social conditions.
Basically, they’re talking about taking into account what your friends and connections are doing, and augmenting your search results based on it.
Before we move on, there is a term used frequently in the patent which could use some clarification. They use the word “restricts” often in reference to both authors and queries. In this context, they are referring to the restricting of the referenced/ranked content based on either the authors/social connections and queries to those impacted by a social condition.
To put it in a more everyday application, using Google Maps would “restrict” the results to those companies/sites where Google knows a physical location for them.
They wrote: … adding a respective author-based query restrict to the set of author-based query restricts; actions further include: identifying a plurality of prominent users that the user is connected … for each of the one or more prominent users in the set of prominent users, adding a respective author-based query restrict … receiving a plurality of scores, each score being associated with a respective prominent user … the score includes a prominence score that reflects a relative prominence of a respective prominent user within the one or more computer-implemented services; the score is based on a number of followers of the respective prominent user within the one or more computer-implemented services … the author-based document restrict is associated with the prominent user … the resource being associated with a social search result in the social search results … the author user has at least a threshold number of relationships with other users … the relationships include asymmetric relationships … the user-based query restrict is specific to the user; the user-based document restrict reflects that an author user distributed a resource associated with a respective social search result to the user using one or more computer-implemented services; the one or more social restricts further comprise a user-based query restrict … the one or more social restricts are associated with a social graph of the user, the social graph reflecting relationships between the user and other users of one or more computer-implemented services.
This is probably the most varied section of the patent with a wide range of important points, so let’s cover each worthwhile bit of verbiage in point form:
What it means:
- Google will index social content and determine the prominent users to which the searcher is connected.
- For those prominent users, Google will generate a query restrict (that is, an augmented search result based on that prominent user and their actions).
- For each prominent user, a score will be generated. This score will be based in part on their prominence in the searcher’s social circles, their number of followers, etc.
- The prominent user may need to have a minimum number of followers or global influence to have a good score (think of links in traditional SEO).
- The restricts (or adjustments) will be specific to the person searching.
- The queries may take into account the full social graph of the searcher.
They wrote: … enhancing serving of resources that are published by prominent users using one or more computer-implemented services … Implementations of the present disclosure are further directed to augmenting queries submitted by searching users to discover resources that are published by prominent users … a resource can be identified as a potential social search result when an author-based document restrict associated with the resource matches an author-based query restrict appended to the query.
What it means: In section 16, we see that the goal of the restricts is to provide to the searcher a set of results that enhances or raises the results that prominent users in their social graph (read: people they engage with often or who are otherwise influential) have previously shared, written or otherwise engaged with.
They wrote: … a social graph can refer to a single social graph or multiple interconnected social graphs … Each social graph can include edges to additional individuals or entities at higher degrees of separation from the user.
What it means: What Google is covering here is that they may take into account a single social network or multiple social networks and that the graph may extend past immediate connections to augment search results based on connections outside those that are direct.
They wrote: … user mail or chat contact, direct contacts on social sites … the social graph includes content generated by individuals (e.g., social networking posts, blog posts, reviews) as connections to the user.
What it means: The scope of the social graph is not simply traditional social sites but all things that could be considered social. This includes email, chat programs, blog posts, review sites, etc. One might even think that their own product Allo could be considered a vehicle to gain connection information.
They wrote: … the membership and degree of separation in the social graph is based on other factors, including a frequency of interaction … respective weights can be associated with nodes and/or edges, and a degree of separation between users can be determined based on the weights.
What it means: In section 31, we read Google’s recognition of making adjustments to the restricts based on factors such as the degree of separation from the searcher (i.e. how closely they are acquainted) and the frequency with which they interact.
They wrote: … the user’s profile also identifies other aliases used by the user … For example, a user may have a first identity for a chat application and a second identity for a restaurant review web site. These two identities can be linked together in order to unify the content associated with that user.
What it means: This section doesn’t really need translation; Google will try to connect different user profiles and, if they can, they will use this connection to add further weight to the searcher’s interests and social graph.
They wrote: Identified content associated with the user’s social graph can include, for example, content or posting to resources subscribed to by the user (e.g., particular blogs). The identified information can also include content generated by members of the user’s social graph. For example, members of a user’s social graph can generate content including, for example, local reviews (e.g., for restaurants or services), video reviews and ratings, product reviews, book reviews, blog comments, news comments, maps, public web annotations, public documents, streaming updates, photos and photo albums.
What it means: In this section, we see the capability to augment search results not just based on what is shared or engaged with by members of a searcher’s social graph, but also what is created by them. They use examples such as restaurant reviews and long-form content like blogs and photos.
They wrote: In some implementations, the presentation and ranking of search results associated with the user’s social graph is adjusted by one or more factors including one or more social signals. For example, affinity can be used to determine whether to show content from a particular member of the user’s social graph or whether to promote or demote the member’s ranking. Affinity identifies the closeness of a member to the user. For example, a friend of a friend who has five common middle friends with the user has a higher affinity than a friend of a friend who has only one common middle friend.
What it means: Here we learn that the patent includes the ability to assign affinity values to the relationship between a searcher and a prominent user. This is to say, the more friends they have in common, the higher the affinity score and the more likely their content would be to impact the other’s search results.
While everything above has been very interesting, it’s in section 51 that we start to see the full impact.
They wrote: … when interleaving search results associated with the user’s social graph along with general search results, a promotion can be applied to the search results associated with the user’s social graph in order to increase their visibility. For example, the ranking of search results associated with the user’s friends is often lower than a general wide-spread result. Thus, promotion of search results associated with the user’s social graph can prevent them from being buried by general search results.
What it means: Here’s where we read the actual intent, and it’s in the line, “promotion of search results associated with the user’s social graph can prevent them from being buried by general search results.”
Where traditionally certain web pages may not naturally rank highly, if prominent members of a searcher’s social graph have engaged with them, they would be promoted to higher positions.
They wrote: The boosting factor could be based on, for example, the number of friends who endorsed the identified resource or a top affinity to a friend who endorsed the identified resource. Boosting can also be based on authorship (e.g., what is the relationship or affinity with the individual that endorsed the resource), or the type of endorsement did the member of the user’s social graph provide (e.g., an explicit endorsement by starring a result or page or an implicit endorsement by visiting the resource or commenting on a posting).
What it means: Where section 51 covered that resources can be boosted based on how one’s social graph interacts with it, in section 52 we get a few more specifics regarding possible methods.
For example, they discuss boosting a resource based on the searcher’s relationship with the author or person that endorsed the resource and the type of endorsement (which seems to add more weight to types that require a great engagement and commitment of time).
They wrote: The search results can be presented to the user, e.g., as a search results page, that includes one or more of the general search results and the search results associated with the user’s social graph. The search results can be presented with separate portions displaying general search results and social graph results …
What it means: In section 55, we read how the search engine results page may include a separate section for content boosted from the searcher’s social graph or those results may be mixed in with the general search results.
They wrote: … a user can be categorized as a prominent user based on a number of contacts that the user has within one or more computer-implemented services, a number of resources distributed by the user as an author user through one or more computer-implemented services, and/or traffic generated by the user within one or more computer-implemented services. In some examples, traffic can be determined based on a number of resources distributed by the author user and a number of follower users that follow the user within the one or more computer-implemented services (discussed in further detail below). In some examples, a prominent user can include a user having a number of asymmetric contacts within the one or more computer-implemented services that exceeds a threshold number of asymmetric contacts. By way of example, the author user can include a famous user (e.g., a well-known blogger, a celebrity, a professional athlete, a head of state, etc.) that has a significant number of asymmetric contacts …
What it means: Here we read how prominent users may be ranked. Prominent users may be selected or prioritized based on a wide array of factors, ranging from the number of people they influence, the amount of traffic they generate, the number of non-reciprocal connections they have and whether they are famous as well as a host of other possible criteria.
They wrote: If the user interacts with content distributed by a second prominent user less often, a second affinity score between the user and the second prominent user can be associated with a second value …
What it means: As always Google allows for on-the-fly adjustments based on user behavior. In this section they cover the issue of promoting content from a prominent user in one’s social graph where the boost is not interesting to one in the search environment.
Consider a scenario where one of your relatives whom you communicate with regularly via social media or email has their content boosted for your queries. If you are looking up election information and their political views are diametrically opposed to your own this would not be at all useful. You may like them, you may engage with them often but that doesn’t make them relevant to your everyday search experience.
So you’ve made it this far, congratulations only a few more points to go. Let’s boil down what this patent includes:
- A searcher can have a social graph associated with them
- This social graph can be created by accessing their social connections including social media sites, blogs, email and chat functions, etc.
- When Google can make a connection between a users persona on one property and another they will to create a fuller picture of their social graph and connections
- Members of a searcher’s social graph can be ranked based on how close they are to the searcher, how many shared connections they have, how popular they are globally, how often the searcher and that user engage with each other, etc.
- Resources by or recommended by a member of one’s social graph can be boosted into a higher search engine result for a query based on the searcher’s association with the user recommending said resource
- If a searcher doesn’t engage with the recommended resources of a specific users consistently, that user’s recommendations may be devalued
In short, according to this patent, what people you’re connected to recommend, like and engage with could be used to impact your rankings.
While I haven’t yet seen any direct evidence of this at this time it makes sense and Google has toyed with similar systems in the past. The idea of boosting a restaurant that a friend of mine likes who lives in a city I’m visiting would have easily perceived advantages.
Similarly, showing me the news sites my friends are referencing on Facebook when I search for election information also makes sense knowing that my political views are more likely to be aligned with the majority of my connections.
There are a significant number of advantages to a system like this, and because it’s foundation is based on engagement it isn’t easily gamed.
I’m pretty sure we’re going to see this type of augmentation of the general search results in the near future taking personalization to the next level. And so continues the journey down the long road of personalized results where now not only will you need a strong site with great content, you need the friends of your prospective clients to engage with it.
Some opinions expressed in this article may be those of a guest author and not necessarily Search Engine Land. Staff authors are listed here.