I wrote this blog because I couldn’t find a good clear answer on how Google uses social signals in their search algorithm. This post looks at the evidence so you can decide for yourself. The use of social signals by search is one of the most heavily contested subjects in SEO.
On one side you have people swearing blind that social has benefited them, a few case studies that indicate social signals do benefit SEO and some correlation studies which imply a strong correlation between search and social media.
On the other hand you have Google saying quite cryptically at many different times that they don’t use Facebook likes, Tweets or Google +1s “at this point in time” (this point in time being between 2011-2014 in the case of this blog – more on that later).
A lot of our knowledge of what drives SEO comes from correlation case studies. Last week I wrote a full strategy guide to search engine optimisation in 2015. I spent a lot of time researching and looking at these case studies. What I learned writing this article is that nothing related to SEO can be considered known, or taken for granted just because somebody says it’s so.
The first three studies I am going to look at today come from 2013 when this kind of correlation study was fairly novel. I believe the first one of it’s kind was done by Moz in 2011, and we can expect one in 2015 which may be illuminating. Correlation studies are seen as a popular way of working out ranking factors. The problem is that all these studies take something that we think is known – ‘that social signals are a ranking factor’ – and testing whether having lots of social signals correlates with higher rankings.
Throughout this post I use ‘old sources’ which are dated by one or two years. Normally I wouldn’t use anything which isn’t fresh or new, but there is a clear trend. If you know your history you have a much clearer idea of where you are going.
This study has a logical progression to it, I strongly suggest you read this post in order. For ease of your navigation I have included a contents.
- Correlations: What Can We Learn?
- Social Signals and Google’s Official Line
- How Google Could Use G+ For Search?
- What Social Signals Are Google Likely To Use?
- More possibilities
Correlations: What can we learn?
Social Signals – anything from a like on Facebook, a pin on Pintrest, a Share, Retweet and all that jazz. Bing objectively uses social signals as it makes their algorithm more accurate. Google have in the past but are now giving very mixed signals. Whats the truth?
These studies look at different (large) groups of SERPs. The first interesting thing that we can tell is that all three case studies have used similar methods to identify whether social signals are a ranking factor, but performed the study on different groups. The overall results, while different, are actually very similar on closer inspection.
First I’ll let you read through the data the three studies have come up with to let you come to your own decision on the answer. One thing that I would like to show through this investigation is that the SEO world is full of bad information coming from misreading reliable sources. SEO is down to testing things and seeing what works for you.
These study’s all score correlation from 0-1, where 0 is nothing in common and 1 is a perfect match. Moz and Searchmetrics both use the Spearman Correlation. Netmark uses The Kendal Correlation (for the more statistical minded among you) .
Netmark’s study – seen above – is the only study which looked at how many social signals were pointed at a domain as well how many likes, +s and shares the page managed to generate. This is very useful because as you can see, there is a much stronger correlation between a pages rank on Google and it’s social interactions than there is to it’s domain’s social engagement.
All three studies found Facebook shares to have a slightly higher correlation with Google rankings than comments and likes.
All three studies found that Google+1s have an edge over the other social media platforms measured. Moz found only a slight statistical difference between Google+1s and Facebook shares, while the other two studies found a stronger difference. The studies also found number of Tweets to have the lowest correlation to search rankings in 2013.
What do these studies tell us?
Looking at three major studies, admittedly from two years ago and finding that they all came out with the same result you could be forgiven in thinking that social shares were a hugely strong – or even the strongest ranking factor. Unfortunately correlation doesn’t prove causation.
Search Metrics found that the number of backlinks a domain has very closely correlates with how many social interactions a page has. Moz found a statistically similar correlation and so did Netmark.
This is very interesting because it is likely if you have a high number of backlinks to your page, you would expect likewise to have a high number of social shares. With all three studies, the ranking factors they tested relating to backlinks match so closely with social signals’ that it is impossible to say that they aren’t measuring the same thing.
A page with thousands of social shares is likely to have had tens of thousands of visitors. If a webpage has tens of thousands of visitors it is more likely to be a good reliable resource which is worth linking to.
Simply, the more shares a page gets the more people view it, the more people who view it the more chance somebody has of thinking that the page is worth linking to. This explains the neat correlation between social shares and links just as well as deciding that social shares are a direct Google ranking factor, although there is a lot more to the social signal debate than just these statistics.
One of the few things that we know about how Google’s algorithm works is that backlinks are a very strong ranking factor. From a social point of view looking at all this information together all you can really tell is that Google+ correlates more strongly than the other social networks across the three studies.
This makes sense as Google+ is owned by the search giant and Google can use data from their own social network far easier than Facebook, where a lot of profiles are private and can’t be crawled. Here’s where it all gets a bit weird.
Why correlation studies aren’t proof
I just want to make a quick point. Onsite SEO your title is considered one of the strongest ranking factors. Search metrics found it to have a 2% correlation with high ranking on Google, which is just wrong – anybody with basic SEO experience knows that.
Interestingly they found the position of keywords in title has a 10% correlation with high rankings on Google. As Search Metrics are testing what correlates with high rankings if all websites they tested had a keyword in their title the result would be a 0% correlation with high rankings.
Without getting to far into the maths this really highlights the importance of using the terms you want to appear for in your page title. They found a 10% correlation with position keyword in title. This not only contradicts the previous 2% correlation but is interesting to a SEO nerd like me – that’s for another post.
Google’s on off relationship with Twitter
Google used to use social signals from Twitter as a ranking factor up until 2011 (source) – exactly how Google used Twitter we don’t know. We do know however that they stopped using Twitter and Google+ launched in 2011.
Google haven’t officially said whether retweets are a ranking factor or not. Google came to a new deal with Twitter at the beginning of this year (2015) which will regain them access to Twitter’s data mine (source). Google confirmed that their deal would start effecting rankings towards the end of the first half of 2015 – that’s pretty much now.
Up until the 2015 deal was struck Google were unable to crawl Tweets in real time because the amount of requests that would generate would cause Twitter server issues. In early 2015 a case study found that at most 7% of Tweets were being indexed.
Google claim that they were were wary of using direct search signals from Twitter because they were cut off “at one point” (see the video in the next section). With the 2015 deal they are able to access the data directly and crawl all tweets in real time.
As covered in the last section, we would expect a post with a lot of Tweets to rank higher than one with nobody Tweeting about it, because this is more likely to be good content that somebody would consider worth linking to.
What is interesting is this is pretty objective proof that Google didn’t use signals from Twitter as a ranking factor in 2013 when these correlation case studies were made. In all the studies it is fairly clear that Twitter was seen as having the lowest correlation with Google ranking out of the three major social media platforms – but Twitter was still considered to have a statistically relevant correlation with search ranking.
Social Signals and Google’s official line
We can learn a lot directly from Google, they have their webmaster blog and YouTube channel. They also interact a lot with people on their forums. Last year Google gave an “official” answer to how they use social signals in their algorithm. It is interesting that Matt Cutts – Google’s head of anti spam – omits Google+ from the following video (published Jan 2014):
Facebook and twitter pages are treated like any other pages on our web index… this many likes on Facebook, to the best of my knowledge we don’t use that [information]… It’s probably because [their content is] really awesome and because it’s really awesome people like it on Facebook.
Now all Matt has actually said here regarding Twitter is that they don’t use the number of followers you have as a ranking signal. But one of our strongest correlating ranking factors – Facebook likes – has apparently been debunked by Google. According to Matt it isn’t a ranking factor at all – it just appears that way.
Matt Cutts tends to be very careful with what he says, listening to what he doesn’t say throughout this video is interesting. He has in the past said that Google “doesn’t use much data from Facebook”.
That would imply that they do use some data from Facebook in their rankings, backed up by the previous correlation graphs showing Facebook as a stronger ranking factor than Twitter. Matt also says that Google crawl Facebook and Twitter like any other website. If that’s is correct, Twitter is public while a lot of Facebook is private so I would expect that number of Tweets would have a higher correlation to search rank than Facebook shares.
Newer data – Similar results
At this point as we have looked at a bit of the history of the debate and other correlation studies I will bring in Search Metric’s 2014 study, compiled after Matt’s video. From this we can see that in 2014 number Facebook shares for whatever reason were still a stronger ranking factor than Tweets.
Both Facebook interaction and Tweets have stayed ranked almost the same as the results from a year before. +1s have dropped significantly and are closer to the results from Netmark and Moz.
Matt Cutts is Google’s head of anti spam, as such he has a tendency to try and make people do what Google want rather than what is necessarily best for their marketing strategy. That said he probably has the best knowledge of Google’s algorithm of any source we can possibly use.
The message I would take from that video is that Google couldn’t reliably trust Facebook – or Twitter up until now – to give them reliable access to their data so it is unlikely that shares or likes have any direct effect on SEO. Most likely this is pure correlation.
Personally I think Facebook probably doesn’t give any direct social signals to Google which are used in ranking, but I can’t completely discount the idea due to the slightly higher statistical weight given to Facebook over Twitter in all four case studies.
Matt has also debunked +1s as a ranking factor.
If you make compelling content, people will link to it, like it, share it on Facebook, +1 it, etc. But that doesn’t mean that Google is using those signals in our ranking.
Rather than chasing +1s of content, your time is much better spent making great content.
Suffice it to say that I would be very skeptical of anyone who claimed that more +1s led to a higher search ranking in Google’s web results
These are three different posts made by Matt on Ycombinator (also 2013). From what he is saying I would guess that Google doesn’t currently use +1s in their ranking algorithms, however haven’t discounted the possibility they may in the future.
This would explain why Matt is being so careful in his wording – he doesn’t want to be quoted out of context on this. The source is 2 years old so it isn’t impossible that they are now using +1s as a ranking factor.
From what Mr Cutts doesn’t say though I would say that it is more likely that number of shares on Google+ is a ranking factor – he is not saying here that they don’t use social signals, or Google+ to help rank websites.
We now have new correlation studies, both from Moz and SearchMetrics. Social ranking factors haven’t changed in any statistically significant way in either study. I was expecting Twitter to become a stronger signal but there hasn’t been a strong change.
One explanation could be the studies’ data was collected before Google had regained access to Twitter’s datastream, or before Google had time to do anything with the data. Either way my findings are just as relevant in 2016 as a year ago when this was first blogged.
How Google could use G+ for Search?
What has been completely ignored up until now, by the three case studies and the two resources that Matt Cutts gives us, is the use of Google reshares. Up until recently (within the last year) sharing a link on Google+ passed over link juice as these links didn’t have nofollow attributed to them.
Now Google does nofollow these links, but Google+ is their network and they can use any information they want in any way they want. Google have said in the past that they “usually” don’t follow links with the nofollow attribute. Could it be that sharing through Google+ is a ranking factor, while +1s and follower numbers aren’t?
A post with a lot of +1s has most likely been seen by a lot of people and considered a good post. This post is more likely to be shared, in turn passing on more link juice. This again neatly explains why Google+ was seen to correlate more highly with high rankings than other social networks in all our correlation studies.
Or maybe Google consider that their rankings were more accurate using Twitter’s data, and now they have access to Twitter’s data again they have stopped following the links from Google+. It’s more likely that now they use certain signals from both networks as ranking factors.
What Social Signals are Google Likely to use?
So Google have used social shares as a ranking factor in the past – that is a known. I would predict that when this year’s correlation studies are done that Twitter will now at least be an equal ranking factor as Facebook, maybe even equal to Google +1s.
If you look at what we know about Google’s algorithm – it is all about classifying the internet in as relevant a way as possible. Google use ‘quality’ as a ranking factor for the backlinks pointing a page. It would make sense that quality is a factor in how they use social signals.
Another clear benefit of using Google+ over other social media platforms is that Google personalises their search results. Your posts often appear prominently to people following you on Google’s SERPs. This can can drive lots of traffic to your website if you have made a useful contribution appearing on a competitive search term.
If you are trying to get a page retweeted or shared, find people are in your industry who regularly share work and are generally seen as a reliable source. Authoritative people are more likely to have a good following and be able to send your work viral, generating you more backlinks and compounding the effect of the social signal alone.
Google classes Reddit as social media rather than a forum or regular website (in Google analytics at least). Reddit works on up and downvoting. People post links (or other content) which start off nofollowed. If the content gains enough upvotes the links become dofollow. When posting your links you can choose your anchor text.
Links from regular forums – even high authority ones – don’t appear to have a powerful practical impact on rankings. The fact that Google classes Reddit apart from a regular forum, and that with enough positive social signals your links become followed I expect this is a signal that Google takes into account.
That’s not to say that they take social signals into account directly, or even directly take Reddit upvotes into account. This is likely what Matt Cutts meant about treating links from social media like any other link. A followed link from Reddit will be on a relevant subreddit with relevant anchor text. Reddit is a high authority website which is geared to pass authority on to relevant content.
For this post to be complete I have to look at hypothetical social signals. Signals that fit with the correlation data and haven’t been flat out denied by anyone who works at Google.
While that doesn’t mean that these are likely ranking factors, they have been raised as possibilities in the past. Google have got a very complex algorithm, so I‘ll let you form your own opinion on these:
- Speed of social reshares and +1s – the speed that people share your content, or even give it +1s or ReTweet could be used by Google for ranking. If something is viral it is relevant and more likely to be searched. It would also be a good reason for Google to setup a new deal with Twitter. They are the fastest moving, most viral social network.
- The wording you use to share could be measured – Google is very good at gauging context. They use a system called latent semantic indexing(LSI) to understand the context of a page. It isn’t so far-fetched to think that the context of a social share might be used to help rank a page – specially if the ‘quality’ of the source sharing your work is used as has been shown with Twitter before Google regained access to Twitter’s data.
- Social engagement is measured – Google could potentially be measuring how many times people reply to your posts on Google+. This is unlikely because people replying to posts you make but not resharing your work is more a sign of controversy than quality – however there is a more likely scenario under the next heading.
These are all would explain why Google+ came out as the strongest social ranking factor by far in five correlation studies. This should give you an idea of just how contested the use of social signals are, and how little we know about how Google works.
It is important to remember that just because something is possible doesn’t make it likely. This section was added to help you make your own decision on what information Google may or may not use – I have my own opinions which I shared earlier in this post but because we are dealing with unknowns I could be wrong.
Find a strategy that works for you, and stick to it as long as you aren’t breaking any rules.
Is the future social?
A recent discussion started by Randy Milanovic looked at whether comments on Google+ could be used to make search results more accurate. Google has over the last year become much better at gauging the context of a page – with their hummingbird update(s) giving Google a better understanding of semantics.
Google+ is their own platform it seems likely that if they aren’t currently looking at comments in discussions to make their results more accurate it is probably only a matter of time. If you use Google+ correctly, you will benefit from improved traffic and you may get a direct SEO boost – what’s to lose?
If you are new to Google+ there is a very effective way of using their circle system to improve your rankings on your followers search results – being logged into G+ gives personalised search rankings.
G+ was most likely created to make the search results more accurate as well as to create an alternative platform to Facebook. It is interesting to think – if Google were able to crawl all of Facebook and Twitter would Google+ ever have been created.
If you have any more possible social ranking factors, which fit with the correlation study, please leave a comment below.
I was unable to find any correlation studies from 2015. Looking at changes year on year or multiple correlation studies can teach us far more than one stand alone correlation study. Correlation studies are notoriously unscientific – you really can’t prove anything with them alone.
Another thing that I wanted this post to highlight is not to listen to just one person when it comes to SEO. Do tests yourself, look at tests other people are doing. Always try to find holes in other people’s studies. People learn by making mistakes.
I am currently speaking with a statistician to see how we might disprove the bullet points in the last section, you can’t prove an unknown but you can disprove them. I will follow up this post as and when more correlation studies occur, and if all goes well may be doing some smaller ones of my own. Subscribe to this blog by joining the email list at the top of the page if you are interested.
Obviously the data I cite is all very old in SEO terms. What is certain is Google uses (or will very soon) use some information from Twitter – otherwise they wouldn’t have made their deal this year. Whether or not this is making any difference to organic results (beyond some mobile rankings having tweets appearing in them) remains to be seen.
Also up until very recently Google reshares passed link juice and affected SEO in that way. Google reshares aren’t the subject of any study, or official resource I could find.
I would guess that reshares probably are still strong ranking signals, even though now they are nofollowed – Matt Cutts has implied Google sometimes ignore nofollow signals.
The obvious correlation between +s and shares, along with the fact that Google has never denied reshares as a ranking factor, although been very vocal on denying other social ranking factors is enough proof for me.
I hope if you take one thing from this blog it is that correlation doesn’t equal causation. If you need that reinforced one more time:
If you liked this post you may also like: