We Analyzed 1 Million Google Search Results. Here’s What We Learned About SEO

We recently analyzed 1 million Google search results to answer the question:

Which factors correlate with first page search engine rankings?

We looked at content. We looked at backlinks. We even looked at site speed.

With the help of Eric Van Buskirk and our data partners1, we uncovered some interesting findings.

And today I’m going to share what we found with you.

Here is a Summary of Our Key Findings:

1. Backlinks remain an extremely important Google ranking factor. We found the number of domains linking to a page correlated with rankings more than any other factor.

2. Our data also shows that a site’s overall link authority (as measured by Ahrefs Domain Rating) strongly correlates with higher rankings.

3. We discovered that content rated as “topically relevant” (via MarketMuse), significantly outperformed content that didn’t cover a topic in-depth. Therefore, publishing focused content that covers a single topic may help with rankings.

4. Based on SERP data from SEMRush, we found that longer content tends to rank higher in Google’s search results. The average Google first page result contains 1,890 words.

5. HTTPS had a reasonably strong correlation with first page Google rankings. This wasn’t surprising as Google has confirmed HTTPS as a ranking signal.

6. Despite the buzz around Schema, our data shows that use of Schema markup doesn’t correlate with higher rankings.

7. Content with at least one image significantly outperformed content without any images. However, we didn’t find that adding additional images influenced rankings.

8. We found a very small relationship between title tag keyword optimization and ranking. This correlation was significantly smaller than we expected, which may reflect Google’s move to Semantic Search.

9. Site speed matters. Based on data from Alexa, pages on fast-loading sites rank significantly higher than pages on slow-loading sites.

10. Despite Google’s many Penguin updates, exact match anchor text appears to have a strong influence on rankings.

11. Using data from SimilarWeb, we found that low bounce rate was associated with higher Google rankings.

We have detailed data and information of our findings below.

The Number of Referring Domains Has a Very Strong Influence on Rankings

You may have heard that getting backlinks from the same domain has diminishing returns.

In other words, it’s better to get 10 links from 10 different sites than 10 links from the same domain.

According to our analysis, this appears to be the case. We found that domain diversity has a substantial impact on rankings.

05_Number of Referring Domains_line

Google wants to see several different sites endorsing your page. And the more domains that link to you, the more endorsements you have in the eyes of Google.

In fact, the number of unique referring domains was the strongest correlation in our entire study.

Key Takeaway: Getting links from a diverse group of domains is extremely important for SEO.

Authoritative Domains Tend to Rank Higher in Google’s Search Results

Not surprisingly, we found that a website’s overall link authority (measured using Ahrefs Domain Rating) was strongly tied to Google rankings:

08_Domain Link Authority (AHREFs Domain Rating)_line

In fact, a website’s overall authority had a stronger correlation to rankings than the authority of the page.

In other words, the domain that your page lives on is more important than the page itself.

Key Takeaway: Increasing the number of links to your site may improve rankings for other pages on your site.

Publishing Comprehensive, In-Depth Topical Content May Improve Rankings

In the early days of SEO, Google would determine a page’s topic by looking strictly at the keywords that appeared on the page.

If the keyword appeared on the page X number of times, Google would determine that the page was about that keyword. Today, thanks largely to the Hummingbird Algorithm, Google now understands the topic of every page.

For example, when you search for “who was the director of back to the future”…

google search for hummingbird

…Google doesn’t look for pages that contain the keyword “who was the director of Back to the Future”.

Instead, it understands the meaning of the question, and provides an answer:

google knowledge graph

As you might expect, this has a significant impact on how we optimize our content for SEO. In theory, Google should prefer content that covers a single topic in-depth.

But does the data agree with that assumption?

To find out we used MarketMuse to analyze 10,000 of the URLs from our data set for “Topical Authority”.

And we discovered that comprehensive content significantly outperformed shallow content.

07_Content Topic Authority (MarketMuse Data)_line

This is interesting. But how do you write content that Google considers comprehensive?

Let’s look at two examples from our data set to find out.

First, we have this article on the Daily Press about the Busch Gardens fun card:

example of page with low topical authority

This page has many of the traditional metrics that result in first page rankings. For example, the page uses the keyword in the title tag and the H1 tag. Also, the domain ( is very authoritative (Ahrefs Domain Rating of 64).

However, this page ranks only #10 for the keyword: “Busch Gardens fun card”.

google ranking number 10 on first page

This low ranking is partly due to the fact the content on the page has a very low Topical Authority score.

On the flip side, we have this page about making Balinese satay sauce.

comprehensive topic content

This page provides a wealth of information on satay sauce. This piece of content covers the history of satay sauce in Indonesia, how the sauce is used, a recipe, and even provides nutrition facts.

Even though this page doesn’t use the term “Indonesian Satay Sauce” anywhere on the page, it ranks on the first page for that keyword:

google hummingbird ranking

Part of the explanation for that ranking is that this page has a high Topical Authority for the topic: “Indonesian Satay Sauce”.

Key Takeaway: Writing comprehensive, in-depth content can help you rank higher in Google.

Long-Form Ranks Higher in Google’s Search Results Than Short-Form Content

Does long-form content outperform short, 200-word blog posts?

We turned to our data set to find out.

After removing outliers from our data (pages that contained fewer than 51 words and more than 9999 words), we discovered that pages with longer content ranked significantly better than short content.

02_Content Total Word Count_line

In fact, the average word count of a Google first page result is 1,890 words.

Several other search engine ranking factors studies have found that longer content performed better in Google.

This correlation could be due to the fact that longer content generates significantly more social shares. Or it could be an inherent preference in Google for longer articles.

Another theory is that longer content boosts your page’s topical relevancy, which gives Google a deeper understanding of your content’s topic.

Also, long-form content’s ranking advantage could simply reflect site owners that care about publishing excellent content. This being a correlation study, it’s impossible for us to pinpoint why longer content performs so well in terms of search engine rankings.

However, when you combine our data with what’s already out there, it paints a clear picture that long-form content is best for SEO.

Key Takeaway: Long-form content ranks higher in Google’s search results than short-form content. The average word count of a Google first page result is 1,890 words.

HTTPS is Moderately Correlated with Higher Rankings

Last year Google called on webmasters to switch their sites over to secure HTTPS. They even called HTTPS a “ranking signal“.

What does our data say?

Although not a super-strong correlation, we did find that HTTPS correlated with higher rankings on Google’s first page.

Use of HTTPS_line

Does this mean you should make the switch to HTTPS today? Obviously, the decision is yours. But switching your site to HTTPS is a serious project that can cause serious technical headaches.

Before you make the plunge to HTTPS, check out these guidelines from Google.

Key Takeaway: Because the association between HTTPS and ranking wasn’t especially strong — and the fact that switching to HTTPS is a resource-intensive project — we don’t recommend switching to HTTPS solely for SEO. But if you’re launching a new site, you want to have HTTPS in place on day one.

There is No Correlation Between Schema Markup and Rankings

There’s been a lot of buzz about Schema markup and SEO.

The theory goes something like this:

Schema markup gives search engines a better understanding of what your content means. This deeper understanding will encourage them to show your site to more people.

For example, you can use the <name> structured data tag to let Google know that when you use the word “Star Wars”, you’re referring to the original movie title…not the franchise in general:

schema markup example 2

Or you can use Schema to show ratings for products on your ecommerce site:

schema star ratings

All of these things should help with your rankings. In fact, Google’s John Mueller hinted that they might use structured data as a ranking signal in the future.

However, according to our analysis, the presence of structured data had no relationship with Google rankings.

Presence of Schema Markup

Key Takeaway: Feel free to use structured data on your site. But don’t expect it to have an impact on your rankings.

Shorter URLs Tend to Rank Better than Long URLs

I typically recommended that people use short URLs for the sake of better on-page SEO.


There are two reasons:

First, a short URL like is easier for Google to understand than

In fact, according to Google’s Matt Cutts, after 5 words in your URL:

“[Google] algorithms typically will just weight those words less and just not give you as much credit.”

And our data supports the use of shorter URLs.

URL Length_line

Fortunately, this guideline is easy to put into practice. Whenever you publish a new piece of content, make the URL short and sweet.

If you use WordPress, you can set your permalink structure to “post name”:

wordpress URL permalinks

Then, whenever you write a post, modify the URL to include a few words:

changing the url

Quick word of warning: make sure the new permalinks only apply to future posts. If you change the permalinks for older posts it can cause serious SEO-related issues.

For example, the URL for my post: 21 Actionable SEO Techniques You Can Use Right Now is simply my target keyword:

google url

Second, a long URL tends to point to a page that’s several clicks from the homepage. That usually means that there’s less authority flowing to that page. Less authority means lower rankings.

For example, this URL to an iPad product page on represents a page that’s far removed from the site’s authoritative homepage:

long url

Key Takeaway: Use short URLs whenever possible as they may give Google a better understanding about your page’s true topic.

Content With At Least One Image Ranks Higher Than Content That Lacks an Image
(But Using Lots of Images Doesn’t Make a Difference)

Industry studies have found that image-rich pages tend to generate more total views and social shares.

This suggests that including lots of images in your content can boost shares, which should therefore improve Google rankings.

To measure the impact of image use on rankings we looked at the presence or absence of an image in the body of the page (in other words, in the content of the page).

According to our data, using at least one image in your content is significantly better than having no image at all.

Content Contains At Least 1 Image_line

However, when we looked at the link between the total number of images and rankings, we didn’t find any correlation.

This suggests that there’s a point of diminishing returns when it comes to image usage and rankings.

Key Takeaway: Using a single image is clearly better than zero images. Including lots of images doesn’t seem to have an impact on search engine rankings.

Using An (Exact) Keyword in Your Page’s Title Tag Has a Small Correlation With Rankings

Since the early days of search engines the title tag has been (by far) the most important on-page SEO element.

Because your title tag gives people (and search engines) an overview of your page’s overall topic, the words that appear in your title tag have long had a significant impact on rankings.

However, we wanted to see whether or not Google’s move towards Semantic Search has made the title tag any less important.

We found that title tag keyword usage still slightly correlates with rankings. However, it had a much smaller relationship than we anticipated.


This finding suggests that Google doesn’t need to see the exact keyword in your title tag to understand your page’s topic.

For example, here are the top six results for the keyword “list building”.

google top 6 results 1

Note how three of the top six results (including the #1 result) don’t contain the exact keyword “list building” in their title tag.

google top 6 results

This is a reflection of Google moving away from exact keyword usage to Semantic Search.

Key Takeaway: Including your target keyword in your title tag may help with rankings for that keyword. However, because of Semantic Search, the impact doesn’t appear to be nearly as great as it once was.

Pages On Fast-Loading Websites Rank Significantly Higher than Pages On Slow-Loading Websites

Since 2010, Google has used site speed as an official ranking signal.

But we were curious:

How much does site speed impact rankings?

We used Alexa’s domain speed to analyze the median load time of 1 million domains from our data set. In other words, we didn’t directly measure the loading speed of the individual pages in our data set. We simply looked at the average loading speed across the entire domain.

And we found a strong correlation between site speed and Google rankings:

Average Page Load Spead (for URL's domain)_line

Again, this is simply a correlation. Could it be that site owners that optimize for speed also optimize for SEO? Sure.

But having a fast-loading site certainty won’t hurt your SEO. So it makes sense to speed things up.

Key Takeaway: Fast-loading websites are significantly more likely to rank in Google.

More Total Backlinks = Higher Rankings

There’s been a lot of buzz about new ranking signals (like social signals) that search engines use today. Many have even gone on to say that backlinks are becoming less important.

We were curious to see whether or not Google still used the sheer number of backlinks as an algorithmic ranking signal.

To measure this, we used the Ahrefs API to determine the total number of backlinks pointing to each page in our data set.

We found that pages with the highest number total backlinks tended to rank best in Google.


Even though Google continues to add diversity to its algorithm, it appears that backlinks remain a critical ranking signal.

Key Takeaway: Pages with more backlinks tend to rank higher than pages with fewer backlinks.

Google Rankings Are Closely Tied to a Page’s Overall Link Authority

In addition to total backlinks, we wanted to answer the question:

Does a page’s overall authority influence rankings?

Most SEOs agree that backlink quality is just as important as backlink quantity.

In other words, it’s typically better to get a single link from an authoritative page than 100 links from 100 low-quality pages.

And our data supports this:

Webpage Link Authority (Ahrefs URL Rating)_line

According to Ahrefs’s measure of link authority (URL Rating), authoritative pages outrank pages with little link authority. However, this correlation wasn’t as strong as the impact of the total amount of referring domains.

Key Takeaway: The overall link authority of your page matters.

Exact Match Anchor Text Significantly Correlates With Rankings

Since Google released its Penguin update in 2012, many SEO professionals have advised against building backlinks with exact match anchor text. However, several search engine ranking studies have found that anchor text is still important.

That’s why we wanted to investigate whether or not anchor text remained an important ranking signal.

Our research shows that exact match anchor text strongly correlates with rankings.

In the early days of SEO, building backlinks with exact match anchor text was a very effective approach. For example, if you wanted to rank for the keyword “online flower delivery” you would make sure your links had anchor text like this:

example of exact match anchor text

However, Google has likely cracked down on this practice, starting with the initial Penguin update. For that reason, we don’t recommend building links that use exact match anchor text, despite the fact that it appears to have a strong impact on rankings.

Key Takeaway: Backlinks with exact match anchor text robustly correlate with rankings. However, because of the risk in exact match anchor text links, we don’t advise utilizing exact match anchor text as an SEO tactic.

Low Bounce Rates Are Strongly Associated With Higher Google Rankings

Many people in the SEO world have speculated that Google uses “user experience signals” (like bounce rate, time on site and SERP click-through-rate) as ranking factors.

To test this theory, we pulled 100,000 websites from our data set and analyzed them in SimilarWeb.

Specifically, we analyzed three user experience signals: bounce rate, time on site and SERP CTR.

We discovered that websites with low average bounce rates are strongly correlated with higher rankings.


Please keep in mind that we aren’t suggesting that low bounce rates cause higher rankings.

Google may use bounce rate as a ranking signal (although they have previously denied it). Or it may be the fact that high-quality content keeps people more engaged. Therefore lower bounce rate is a byproduct of high-quality content, which Google does measure.

As this is a correlation study, it’s impossible to determine from our data alone.

Key Takeaway: Google may use bounce rate as a ranking signal. Or it may be a case of a correlation not equaling causation.


Special thanks to our data partners: SEMRush, Ahrefs, MarketMuse and SimilarWeb for making this study possible.

I also want to thank Eric Van Buskirk of ClickStream (Project Director) and Qi Zhao (Head Data Scientist) for their contributions.

Also, if you’d like to learn more about how we collected and analyzed our data, here is a link to our study methods.

      1. Good to see some of the data backing up the claims so many folks make, amazing work by you and the other contributors. In the end, it appears that long form quality content and citations are still the most valuable signals.

      2. “4. Based on SERP data from SEMRush, we found that longer content tends to rank higher in Google’s search results. The average Google first page result contains 1,890 words.” –> Does this mean ALL words on the page, including product descriptions etc.?

        1. I’m also curious about this topic. Nearly all websites have navigational links, excerpts of related content, author bios and so on. Together, they can add a significant amount of words to any webpage.

          I’m not disputing that longer articles tend to rank better. We see that every day.

          But did the study measure article lenght or whole page word count?

          1. The study likely measured whole page word count, there are a number of free tools online for counting words on a page and most of them pick up any and every word including navigation, author bios etc. 🙂

    1. Yo Yo Chase.

      Soooooooooooooooo, I’ve got to ask.

      Did any of these make you say “sayyyyyyyyy what”

      For Me:

      When I read “Despite the buzz around Schema, our data shows that use of Schema markup doesn’t correlate with higher rankings.”

      I figured there would be more value with SCHEMA relating back to SEO. I mean other factors can play into too.


      Chris Pontine

        1. Lots of value here, but the schema finding in particular is one where I have to wonder if the industry/vertical may make a difference. After getting to know the search results around a few hundred travel related terms, one of the key patterns to emerge was that 2 of the sites that appeared in the top 3 for a majority of the search terms were also the only 2 sites using schema markup on the core content (not just basics like breadcrumbs or site type). I haven’t done as much research, but I’d imagine recipe sites would show a similar pattern.

          In the travel search case, for all but a few of the terms, external links weren’t much of a factor for anyone on the first page. Internal links, and on page factors (including schema) played a larger role. I don’t think Google has an entirely different algorithm by vertical, but I think there are some where the absence of links – especially to detail pages – is natural and other factors play more of a role.

          1. For sure industry and vertical plays a big part here. Good call. Many sites have content that doesn’t have much to mark-up with schema.

            Also there’s a very concise chart on the method and results section and you’ll see how important each factor is relative to others factors. A “.1” shows exactly 2X as much importance in our study as a “.05”

          2. Schema still has indirect value in many verticals (like travel as you mentioned), since adding the markup will get you star ratings and such displayed in SERPs.

          3. Of course if Schema drives a higher CTR, this gives you more potential for people to see and link to your content, driving rankings. It also potentially means attracting more relevant traffic (e.g., showing prices might deter the wrong audience from visiting), in which case bounce rates might go down. So there are different ways to look at the value here. We should also remember that someone using Schema may be more prone to using other SEO techniques, which is why a larger study like this one is useful for weeding out exceptions.

            Good point on some verticals naturally having fewer / more links on certain types of pages.

        2. The content of the answer box for the question ‘who was the director of Back to the Future’ is often sourced from Schema markup. That’s my understanding of the situation anyway.

        3. I think this article could use be expanded in the discussion around schema. The fact that schema isn’t present in the URLs that rank number 1 doesn’t mean that schema isn’t having an impact on rankings. Would those URLs ranking in the 2-4 slot where you’re seeing schema have done worse without schema? That seems to be the question.

    2. Hey Brian. I’m a reluctant study of SEO. I’m a business owner first and foremost and kinda despise having to handle our SEO. But its just too expensive and confusing to outsource for our 3 person company. Your post and emails have made it possible for me to evaluate what parts of SEO are important to my business right now and which are not. Most SEO’s and web marketers want to sell me a bunch of stuff that I didn’t understand or even need in the past. Now I can converse with people in your world and direct them properly on the projects I wish to outsource in the future. Until now, SEO has been a part of my business I pour money into and get zero results and heaps of excuses from my contractors (14 years and counting) Your info is really helpful. Thanks.

    3. Thank you for this fantastic piece of work. I also find by the way that the keyword in the domain name is still an important ranking factor even though Google mentioned several times its loosing importance. Not as just yet. I have several sites that rank just because the keyword is in the domain. What’s your take on this?

  1. This ROCKS Brian! #shared wow just goes to show you not only need to build authoritative Backlinko but you need good, relevant on-site content for people to stay to make a good blend for success in the SERPs.

      1. Nope, my opinion…

        Darren nailed it!

        I’ve read a few of your posts, and that term should definitely be renamed:

        ‘Authoritative Backlinko – Formerly Known as —>>> (authoritative backlinks)’

        Seriously 😉

      1. Could it be that some of the sites who rank high with exact match anchor text have not yet been penalized?

        I mean what is your protection from skewness and what is your R-squared in this data set?

        I’m not saying you’re wrong, your findings are probably true, however, there may be a possibility that remnants of an old ranking factor still lingers and will not be as beneficial going forward.

        1. That could be, Robin. But with 3+ years of Google Penguin updates, I doubt too many sites that used manipulative anchor text are still around.

          Also, keep in mind there are pages with a lot of legit exact match anchors. So they wouldn’t be penalized.

  2. Fantastic study! I don’t think I ever shared an SEO article on my Facebook but you have won sir. I hate when people obsess over Schema markup. It is interesting to see bounce rate coorelates with higher rankings. I guess it is because those are more high quality pages as well…

    I think the hardest thing for SEO is deciding on anchor text. Yes it helps, but who wants to be the guy who hurts his clients. Even though it helps I am of the mindset to play it safe…

    1. Thanks Ronald.

      I’m with you 100%. Schema is nice to have. Nothing wrong with it. But it won’t magically boost your rankings.

      1. The Schema points are interesting, but not surprising. We’ve used it since day dot and have never seen it boost rankings, we have however seen sharp increases to click through rates for products, thus still reccomend adding it where possible.

  3. Great study!

    #9 is surprising!

    9. Despite Google’s many Penguin updates, exact match anchor text appears to have a strong influence on rankings.

    1. Thanks Mitesh. It looks like it still helps Google understand a page’s topic. They probably didn’t want to throw the baby out with the bath water!

    1. Thanks Rob. Oh man, let’s just say we had a team of people work really, really hard for a really, really long time. To my knowledge that’s the only way to get things done!

    2. As part of the team, I can say Brian could write a novella (are those non fiction as well?) about how this was done. The number of servers, CPUs, scripts, late nights, API calls for data, etc.: this was a project full of big numbers: 1,000,000 results and several data partners was just part of that.

      1. Who pays for all this?

        How does this relatively short article produce a return on the investment required to do the research?

        1. It drives newsletter registrations, builds personal brand, and pushes business to “partners”, which is the whole point!

      1. Even if the HTTPS ranking improvements were worth 5%, it’s still appears as loss for most sites because of the ~10 – 15% authority lost from site wide redirects. Doesn’t equate to a winning proposition, not to mention the potential disasters that can come from a botched implemntation.

    1. They seem to ignore the fact that sites using https might in general, just be higher quality sites which the admins put a lot of time into.
      The fact that they are secure may actually have no influence at all, just that sites with better quality content are more likely to be secure.
      Since the advent of free ssl certs I’ve switched a lot of sites to secure and none of them have seen any change in ranking.

    1. In general, the main ways to reduce bounce rate are to:
      1. Make it easy to skim your content (e.g., add bold, italics, headings, subheadings, lists, images, and so on)
      2. Add more internal links into the text of your articles to allow for users to browse to and find other content on your site
      3. Make sure there is a clear sales/conversion funnel. If you want users to navigate to a sign-up page, make sure it’s clear how to get there. Things like adding a “Get a free trial” button to the sidebar, a call to action at the end of your articles, and so on, are important in terms of user retention and lowering bounce rate.

Leave a Comment

Your email address will not be published. Required fields are marked *