We all know social media activity can boost organic search visibility, acting as signals by which search engines judge content quality. But what are the negative effects of such an interaction, if there are any? Furthermore, if social indicators are used to judge content, why are algorithm updates still spoken of in relation to links? What could social mentions indicate about the domains on which the shared content resides? Are there different effects according to who does the sharing? The purpose of this article is to pose questions which need to be asked, and describe some evidence which lead me to the possibility of negative social citations. Please remember, I am an optimist merely seeking the boundaries of which we will inevitably encounter.
Search and social are an integral part of an inbound marketer’s strategy. Tweets, likes, and especially +1s are all fair game for search engines to judge content by. For example in late 2010, Google said this of retweets:
“Yes, we do use it as a signal. It is used as a signal in our organic and news rankings. We also use it to enhance our news universal by marking how many people shared an article.”
“We take into consideration how often a link has been tweeted or retweeted, as well as the authority of the Twitter users that shared the link.”
When asked “Do you try to calculate the authority of someone who tweets that might be assigned to their Twitter page. Do you try to “know,” if you will, who they are?”
“Yes. We do calculate the authority of someone who tweets. For known public figures or publishers, we do associate them with who they are. (For example, query for Danny Sullivan)”
“Yes we do compute and use author quality. We don’t know who anyone is in real life :-)”
Thus, two dimensions of interest: quantity and quality of shares and sharer- at the heart of Google’s attempt to compute reputation scores and trust. That’s all well and good, and I’ve seen the positive effects of social media on organic search for numerous clients. But a cloud looms on the horizon of happy happy share land, because this form of assessment is incomplete. For example, try to measure the capacity of a glass that is half full, based only on the water it contains. What about the empty space? Would you not measure that too?
First, I want you to envision a spammy social media account. Hold it in your mind- how it works, how it looks and acts, and the systematic shares that happen like clockwork. Now, envision the typical implementation of marketing automation. Is there a difference? Do you think an algorithm can tell the difference?
From Rob Garner’s book, Search and Social (which I highly recommend):
“The spam problem is a major issue that search engines have dealt with since they first began. The fundamental issue for search and social providers delivering results in real-time is that they must trust a content provider to determine whether it is “spammy” or a reliable source of content. Although search engines have gotten very good at studying the signals of trust that point to high-quality content, the imperative for marketers is to cultivate trustworthy websites, content, and a social-network presence that show it is backed by a real content producer, used by real people.”
How might an algorithm determine what is quality content and who are reputable sharers? They have more resources and are capable of much more sophisticated analyses than ourselves, and one thing we can do ourselves is use social APIs. Take just a few sips from the Twitter fire-hose and you’ll soon realize that although the data set is huge, it only contains a handful of fields including date, user, location, tweet content, and link. It’s an easy assumption to make that these are all being analyzed in some form, whether for determining content velocity (for example Top Tweets) or the lack thereof. Keep in mind all social media networks have their own APIs and data types, but at their core all collect basic things like the user, the content, a time-stamp, and the link (if there is one). In other words, every single time a piece of content is shared, data points are created for who, what, where, and when.
This information allows for simple data science procedures to answer a few basic questions about social media events, such as:
- Who or how many people have shared this individual piece of content? Is it one person, a group of people, or a wide variety of people?
- What day of the week or time of day is this piece of content shared the most? Are these times random or is there a discernible pattern?
- What is the semantic context with which the content is shared? Is all of the language the same? The reading level? Is it just the title of the piece or is there variety of the surrounding text when it is shared?
- What content is an individual user sharing? When? Do they share the same piece of content repeatedly? Are they favoring content from any particular domain?
So, what am I getting at? We know speedy & reputable sharing creates fringe benefits like improved organic search visibility. But what about the opposite, the social media miscreants? Specifically, the accounts who only share content from their own web property, on the same schedule, in the same networks, repeatedly.
Put yourself in a search engine’s shoes (if they wore shoes). Over time, patterns form from the same account sharing the same content- and it probably sticks out like a sore thumb. Check out slide 29 of this deck from Bill Slawski about search and social patents. OK, I’ll save you the trouble of flipping through the slides- User Rank factors in “originality in relation to previously-posted content.”
The spontaneous nature of viral content and the benefits it provides the domain on which it lives is undeniable. What I’m suggesting is to consider the opposite effect. We know that social interactions contribute positive signals about content, manifested through organic search- so why not negative signals too? Social signals not only determine the reputation of the account within a network, but likely the domains they are tied to. After all, how many auto-pilot social media accounts are out there that aren’t somehow related to the site they’re promoting? Most social media profiles indicate what domains they represent simply through their bio section, but if the profile itself doesn’t contain the owner’s domain it can be easily assumed from their activity. Google+ gives you the opportunity to list every single profile and site you’re associated with! The idea here is the possibility that social media profiles are being associated with domains.
“Social relevancy and trust is measured on the same model that search engines use to measure the authority and trust of websites, as well as the linked connections between them.”
Recall the legend of the link. Once upon a time, good links were great and bad links were nothing. Then Google decided bad links would count against you, creating the present situation of disavowals, negative SEO, and the need to prune links like a kudzu vine. If you don’t keep on top of this, you could be looking at spending quality time removing a Google penalty. It has gotten so bad that people might even try to remove your good links without you being aware. Negative SEO is linking to your domain from low quality domains, and I propose that negative social citations could work similarly through low quality social profiles.
Although I have no hard data (yet) it makes sense that social signals could end up hurting you just as much as they could help you. So far, the only evidence I do have is from colleagues. This last round of Penguin updates didn’t effect as many sites as some people expected (or hoped), and I only have second-hand information on 4 sites which were hit. These individuals don’t know each other, but I looked at each of these sites and their link profile to try to gain some understanding of what was going on. Sure, there were some low quality links, that was to be expected, but none were so different from the profiles of unaffected sites to draw any conclusions. I started watching their marketing activities in general, and found that all 4 of these businesses were engaging in automation of their social media accounts, sharing their own content through their own branded profiles, on regular, predictable schedules.
“The recommendation here is to avoid using auto-follow services to grow your following. Do it the old fashioned way: share useful content. Your followers will appreciate it and recommend you to others.”
From a search engine perspective, this hypothesis makes a ton of sense. If you too were tasked with finding the best websites with the best content, you would most likely look at how that content is being shared and by whom. But would you only look for positive signals, or would you expedite your efforts by also looking at negative indicators? Classic inductive and deductive reasoning.
My advice is not to stop your marketing automation, but be smart about it. Share more than just your own content (curate don’t dictate), vary your schedule, engage and outreach, and don’t be a broken social media record. When the next algorithm update comes along and you can’t figure out why you slipped, keep these possibilities in mind. I have yet to hear or read about the possibility of negative signals being passed to domains through social media accounts, and as this idea arose organically, I would appreciate any and all responses to this hypothesis.
In the effort to determine what is good content and who is reputable, search engines are increasingly utilizing social signals. However, industry literature only deals with the positive signals of social media’s effect on organic search, and not the negative. My hypothesis is that, like links, social media activity could not only be associated with domains (owners), but also contribute to the algorithmic perception of their quality, both positively and negatively.