It’s the biggest mystery and controversy of Google’s search ranking algorithm. For a long time, the SEO community has debated: is the click-through rate (“CTR”) of search results listings a ranking factor? Or the closely related “bounce rate” and “dwell time”?
I present to you everything Google has ever said about this, along with some observations and opinions.
Table of Contents
Clicks, CTR, bounce rate and dwell time
If you are newer to SEO, the concept of clicks or click-through rate (“CTR”) being ranking factors is simple to explain. Once a user performs a keyword search, they can then click on a listing on Google’s search results page. Google could count those clicks as a type of vote for the content in the results and lend more ranking ability to those listings that draw more clicks for the keyword in question.
Similarly, “dwell time” would be counting how long one stays on a webpage after clicking through to a page from the search results.
A “bounce” happens when one clicks through to a webpage and leaves without navigating to another page. The assumption is that if a bounce happens too rapidly, the user may have found the page’s content unsatisfactory for their query.
“Dwell time” is also how long the user may linger on the webpage before clicking elsewhere or back to the search results. All of these signals center upon the click to listings in the search results.
The mystery: Are CTR and bounce rate ranking factors?
Despite many of my colleagues believing Google’s official line about CTR or bounce rates not being ranking factors, I will confess that I have long wavered on the question, and I have often suspected it indeed could be a ranking factor. In a recent poll I took on Twitter, CTR was voted the most controversial of all ranking factors.
However, there are a lot of good reasons to believe Googlers when they tell you what does or does not influence search rankings. I have worked in information retrieval myself, and I have known and conversed with a number of official Google evangelists in person or via chats, emails, etc. – and they uniformly give great advice and all seem to be highly honest and generally good people.
…there have been those moments when something rises and sticks in rankings that do not seem like it should, based on all the classic ranking factors that we know.
I have long worked in online reputation management where SEO is leveraged heavily to try to improve how a person or organization appears in search when their name is queried.
There have been these weird instances where a nasty blog post or article with few or no major external links will abruptly pop up in the rankings – and, it just stays.
In contrast, other content that has been around longer and has stronger links just cannot gain traction against the nasty-gram item.
You cannot help but notice the difference when these reputation-damaging items arise on the scene. Such pages often have scandalous and intriguing titles, while all the other pages about a subject have more normal, conservative titles.
When you search for a name, and you see some title referencing them along with the word “lawsuit”, “indictment”, “exposed”, “arrested”, “scam”, etc., you are immediately curious, and you will want to click to hear what it is all about.
I have sometimes described this as “rubbernecking on the information super-highway” because it is like how people are drawn to slow down and look when they see a terrible wreck on the road. You see the scandalous title in the search results, and the impulse is to click it.
It has often seemed like the scandalous headlines keep drawing clicks, and this activity seems to buoy the content into appearing high in the rankings on Google’s Page 1.
I have even engineered more scandalous headlines on positive pages to draw attention for a client. Once that engineered content is getting most of the attention, the original negative item starts to subside in the results. When this happens, it seems like users’ clicks are to blame.
But, is the dynamic just coincidental correlation? Or is it exactly what it appears it could be – an outcome based, in part, on quantities of relative click-through numbers?
Get the daily newsletter search marketers rely on.
Reasons to suspect Google uses CTR as a ranking factor
Beyond my anecdotal examples, there are a number of good reasons to suspect that Google could use clicks of links in the search results as a ranking factor. Here are a few:
1. Google has long tracked clicks on its links
If this is unused data, why track the clicks? I tried to recall when I first glanced at Google results’ HTML and saw that the links were being tracked. It might be sometime in the early 2000s.
What do they do with all that data? After the advent of the inclusion of search analytics in Google’s Webmaster Tools (later renamed to Google Search Console), this click data was at least used in webmaster reports.
But, it was collected by Google well before the search analytics report.
2. Google tracks clicks on ads
Click data affects rankings within the paid ads section. So, why wouldn’t they do the same in organic?
It would not be a surprise if Google used a similar method in organic that they use in paid search, because they essentially have done that with their Quality Score.
Over 15 years ago, Google rolled out its Quality Score, which affects ad rankings – and there is now ample evidence of Google using quality criteria in organic rankings.
While different parts of Google – such as keyword search versus Maps – use different ranking methods and criteria, Google sometimes cross-pollinate methods.
3. Google disclosed in 2009 that clicks on search results affect rankings under personalized search
If it is used or has been used in the past for personalized search results, it clearly can be used for regular results, too.
4. An independent researcher examined click-throughs as a ranking factor and found it to be a potentially valuable method
Dr. Thorsten Joachims examined click-throughs as a ranking factor and found it to be a potentially valuable method. Notably, he found:
- “The theoretical results are verified in a controlled experiment. It shows that the method can effectively adapt the retrieval function of a meta-search engine to a particular group of users, outperforming Google in terms of retrieval quality after only a couple of hundred training examples.”
Thus, in a limited study, it was found to be effective. Considering this, why wouldn’t Google use it? Of course, his definitions for “outperforming Google” and determining usefulness likely differ from the criteria used by Google.
5. Bing uses click-throughs and bounce rate as ranking factors
Microsoft Bing search engine confirmed that they use click-throughs and bounce rate as ranking factors. However, they mentioned caveats around it, so some other user engagement context is also used for evaluation.
Search engines certainly use different signals and methods to rank content in search results. But, it is an interesting counterpoint to rhetoric that it is “too noisy” of a signal to be useful. If one search engine can use the signal, the potential is there for another.
6. If Google convinces people that CTR is not a ranking factor, then it reduces Google search as a target for artificial click activity
This makes it seem like there could be a substantial motive to downplay and disavow click activities as ranking factors. A parallel for this is Autocomplete functionality, where users’ searches, and potentially also click activity, used to be very prone to bot manipulation.
Google has long disliked artificial activity, like automated requests made by rank-checking software, and has evolved to detect and discount such activities.
However, bot activity in search results targeting ranking improvement through artificial clicks would likely quickly become more significant than they already handle. This can potentially create a negative impact on services similar to DDoS attacks.
Despite the years and years of stating that CTR is not a ranking factor, I have seen many jobs posted over time on microtask platforms for people to perform keyword searches and click upon specific listings. The statements may not have accomplished deterrence, and Google may already be effectively discounting such manipulation attempts (or they are hopefully keeping some of that artificial activity out of Analytics data).
7. Google AI systems could potentially use CTR and Googlers would not know if or when it was impacting rankings
Three years ago, when I wrote about how Google could be using machine learning to assess quality of webpages, I strongly suggested that user interactions, such as click-through rate, could be incorporated into the machine learning models generated for a quality scoring system.
An aspect of that idea could potentially happen, depending upon how Google builds its ML systems. All potential data points about websites and webpages could be poured into the algorithm. The system could select ranking factors and weight them according to what matches up with human quality rater assessments of search results.
With such massive processing power to assess ranking factors, an algorithm could theoretically decide if CTR was or was not a useful predictor of quality for a particular type of webpage and/or website.
This could produce ranking models for many thousands of different kinds of webpage and search query combinations. In such a system, CTR might be incorporated for ranking scientific papers but not for Viagra product pages, for instance.
The mystery remains
You might think that that third point would essentially set the record straight as Google flat out stated the ranking factor for personalization. But the mystery and controversy remain as the question centers upon overall rankings in a broader sense beyond just personalized results. The controversy surrounds whether CTR is used as a core ranking signal. The blog post disclosing clicks as a personalized ranking factor was from 2009 – when personalization effects seemed a little more overt in search.
Because there is some reasonable basis for thinking Google could use CTR as a ranking factor more broadly beyond limited effect in personalization, it creates the groundwork for many SEOs to easily believe that it is indeed a major ranking factor.
Of course, one of the biggest reasons people in SEO have come to think CTR is a ranking factor is because it naturally has a high correlation with rankings.
This is the high-tech version of the age-old question: which came first – the chicken or the egg?
The links on the first page of search results have the vast majority of clicks for any given query, and on the first page of search results, the higher ranking listings typically receive more clicks than those that are lower. This makes CTR as a ranking factor seductive.
The obvious question is: Is this coincidental correlation or is it evidence of causation?
Where cause and effect are so closely intertwined, the prospect of confirmation bias is very easy – and this is why one should be extremely careful.
This leads us to what Google has said over time about CTR as a ranking factor.
Everything Google has ever said about CTR as a ranking factor
Former Googler Matt Cutts commented that bounce rate was not a ranking factor, stating that it would be spammable and noisy (meaning it would contain a lot of irrelevant data that is unhelpful to ranking determinations).
In a Google Search Central video, Cutts was asked, “Are title and description tags helpful to increase the organic CTR – clicks generated from organic (unpaid) search – which in turn will help in better ranking with a personalized search perspective?”
He only answered a part of the question, saying that “…so many people think about rankings, and stop right there…”, advising the person to improve their page title, URL and snippet text to help their CTR.
He avoided answering whether CTR could affect rankings. Of course, this question was specific to personalized search.
Nine months later, Bryan Horling, a Google Software Engineer, and Matthew Kulick, a Google Product Manager, disclosed that clicks on listings were used in rankings in personalized search, as I noted above.
An FTC Google Probe document (regarding an antitrust evaluation) was leaked to the Wall Street Journal. It recorded a statement from Google’s former chief of search, Udi Manber, saying:
- “The ranking itself is affected by the click data. If we discover that, for a particular query, hypothetically, 80 percent of people click on Result No. 2 and only 10 percent click on Result No. 1, after a while we figure out, well, probably Result 2 is the one people want. So we’ll switch it.”
The document further reported that:
- “Testimony from Sergey Brin and Eric Schmidt confirms that click data is important for many purposes, including, most importantly, providing ‘feedback’ on whether Google’s search algorithms are offering its users high quality results.”
A bit of the context is missing in this document because the segment about rankings and click data comes directly after a missing page – it appears that all the odd pages from the document are missing.
Danny Sullivan, former Editor-in-Chief of Search Engine Land, and current Search Liaison at Google, tweeted about the leaked document’s reference to rankings being affected by click data, stating: