Does Anchor Text Still Influence Rankings? I Analysed 255 Domains To Find The Answer [Case Study]
Most white hat link builders would give you the following advice with regards to anchor text:
Build a mix of branded links and natural, long anchors — keeping exact match links to a minimum.
The argument/theory being that Google is likely to penalise your site for too many exact match anchor links.
But how many is too many?
And anyway, who cares about theory?
Especially when we have the greatest backlink profiler in the world at our disposal(that’s Ahrefs in case you are wondering) and can delve into profitable niches to see exactly what is going on.
So, I decided to take a look at 51 massive (and in many cases notoriously spammy)keywords and examine the link profiles of the top 5 Google results for each.
The results were interesting…
A Quick Introduction To Anchor Text
I’ll keep this brief. For a more detailed guide to link anchors, check out my anchor text post on the Ahrefs blog.
Anchor text is…
And it looks like…
Way back in the late 90s and early 2000s (when SEO was far less complicated than it is today), Google weighted the text used in the anchor pretty heavily in its algorithm.
If you wanted to rank for ‘blue widgets’, then lots of links with the exact text ‘blue widgets’ would soon see you hitting number 1 in the search results.
Of course this was clearly very open to manipulation and as us SEOs just can’t have nice things without hammering them to death, Google began reducing the weight of exact match anchors as a ranking factor.
Indeed, as the years and animal themed updates rolled by, those calves who had particularly fatted themselves on exact match anchors began to see their sites slapped with penalties.
So nowadays, we all strive for a mix of anchor text (distribution) that looks something like this…
And we spend a lot of time worrying about the percentages of each in our profile, and whether we are in danger of crossing that fine line between optimised and over-optimised.
This email, received by Ryan Stewart of Webris from a client he built some solid white hat links for, sums up today’s twitchiness about anchor text.
“We are looking to dilute our anchor text profile. We hired you to build white hat links — anchor texts should ONLY contain our brand name! Anything else is manipulation and we will get penalized by Google.”
So is Ryan’s client correct? Should we really be avoiding using any keyword phrases in our anchor text to avoid penalty?
Well, let’s find out…
I chose to analyse competitive 2–3 word phrases in big, high paying niches — law, finance, insurance, health etc —with high difficulty scores.
I had a fair idea of what keywords to look at (based on my experience with affiliate marketing), but also found a couple of helpful guides on high paying keywords/niches (for example this one from wordstream).
To deep dive into niches for keyword ideas and for data on search volumes/keyword value I used Ahrefs Keywords Explorer.
There is a saying in songwriting “don’t bore us, skip to the chorus”, so I’ll go straight ahead and give you the results of the study, before outlining the process I used to collect and analyse the data.
I primarily wanted to look at keyword rich anchors and their current influence on search, so will be focusing on exact match and phrase match, but there are some other interesting domain/link related stats that I will share too.
Exact Match Anchors
Here are the average and median percentages of exact match anchors by position.
To clarify this is the percentage of exact match anchors from individual referring domains, calculated against the total number of individual referring domains to the URL.
I included the median figure as there were certain niches (for example payday loans)that were very anchor text heavy, so I was conscious they may distort the percentages.
Now for the actual volumes (number) of exact match anchor links.
And finally (for exact match) the distribution of DoFollow and NoFollow exact match anchors (as percentage of referring domains).
The data above would suggest that exact match anchors still have some influence on top placements in competitive niches.
Average is high, but as mentioned the data is skewed somewhat by certain niches, so I would err towards the median results and suggest a percentage of 1 to 2% (and a low number, i.e. 2 or 3) would be safe and potentially help with rankings at page level.
Phrase Match Anchors
Here are the average and median percentages of phrase match anchors by position.
This is where the search phrase appeared in its entirety within the anchor text(including exact match) and again is calculated as a percentage of the total number of referring domains to the URL.
And the volumes…
Again, let’s look at the distribution of DoFollow and Nofollow phrase match anchors as a percentage of the number of referring domains.
The data above would suggest that anchor text containing a target phrase continues to influence rankings.
Average and median figures are closer than with exact match, and it would appear that around 30% of anchors containing a target phrase would be a desirable and safe level.
It would also appear that a fairly even split between DoFollow and NoFollow correlates with rankings.
Number Of Domains With At Least 1 Keyword Anchor Link (Phrase Or Exact)
The following data shows the number of ranking domains containing at least one anchor text link (split by phrase and exact).
If we look at SERP position 1, it becomes clear that there is a correlation with a page’s link profile containing at least one keyword rich anchor and its ability to rank highly for that keyword.
Here are statistics on average Domain Rating, URL Rating and number of referring domains (average and median) for positions 1–5 across the 51 keywords.
It would appear that for a page to rank for a competitive keyword, a domain rating of 60+ and a URL rating of 30+ is desirable.
Number of referring domains is less easy to analyse as clearly not every link is created equally, but for information there were only 6 pages which ranked (out of 255) which did not have any external links pointing at them.
The median number of referring domains across all results and positions was 53.
This would suggest that for competitive keywords, domain strength will not normally be enough for an individual page to rank and link building directly to individual pages may assist with their rankings.
You can check your own site’s Domain Rating and page’s individual URL ratings using Ahrefs Site Explorer.
The data collection ended up being rather time consuming.
In retrospect I may have been better scraping the Google results and writing a script to interface with the Ahrefs API. But I don’t have a time machine to undo the past, so here is what I did…
For each keyword I grabbed the top 5 organic search results from Google UK.
I put the URLs into a spreadsheet and then used Ahrefs Site Explorer to analyse each result — starting by pasting the URL into the search box (with search parameter set to URL).
From the summary page I collected the URL Rating, Domain Rating, and number of Referring Domains and added them to my spreadsheet.
Next I ran the Anchors report (Backlinks > Anchors).
And filtered the report by the search phrase I was analysing.
I collected the number of Dofollow referring domains for the exact match keyword(in this case 13) and also the number of Nofollow referring domains (total referring domains minus Dofollow referring domains — in this case 0).
Next I exported out the data into a spreadsheet to calculate the total number of phrase match anchors (Nofollow and Dofollow).
The calculation for DoFollow was the sum of all DoFollow links minus the number of Exact Match DoFollow (in this case 13).
NoFollow was calculated as the sum of all Referring Domains with the phrase, minus total number of DoFollow phrase match, minus total number of Referring Domains with exact match anchor.
All of this data was added into a big spreadsheet for analysis — a version of which is shared publicly here.
While I tried to be as thorough as possible and include a reasonably large data set, there are two obvious caveats which apply.
Firstly, there is the standard correlation does not imply causation.
Google has over 200 ranking factors, and while the data above would suggest that keyword usage in anchor text continues to be one of them (and you probably should build some keyword rich links) there are numerous other factors which may have influenced the rankings of these pages.
Secondly, I am unable to tell if any of the links I discovered have subsequently been disavowed, which could be the case in some of the particularly keyword rich profiles.
So as always, take everything with a grain of salt and conduct your own tests.
The niche where anchor text seemed to have the biggest influence on rankings was one that is so notoriously spammy that it even had a Google update named after it — payday loans.
In fact many of the ranking pages had upwards of 80% in keyword rich anchors.
Check out the chart below, which shows the top 10 results in this niche.
Conversely, the niche where anchors seemed to have absolutely no influence on rankings (indeed none of the top 5 results had any exact match or phrase match anchors)was another notoriously spammy one to do with male ‘performance’ pills.
Caveats in place, I’m going to stick my neck out and say that keyword prevalence in anchor text is almost certainly still a ranking factor, and perhaps a bigger one than Google would like us to believe.
So, if you’re building links (or someone is building links for you) then you probably don’t need to be too sensitive about anchor text, and a few keyword rich links as part of a diverse profile should help you rank.
Just don’t go crazy though and from the data above, I would suggest exact match at around 1.5% and phrase match at around 33% will keep you on the right side of Google.
Of course, as always, the strongest and safest links are those which are editorially earned (read how I picked up editorial links from the press here) and you should always avoid building low quality links, regardless of anchor text.
After my study, we decided to repeat the process, but this time analyse a whopping 16,000 keywords. You can find out the results here.