How to Unnaturally Naturally Vary Your Anchor Text

by on March 2, 2011 | posted in SEO Theory

Something you’ve undoubtedly heard, if you’ve done SEO, if you’ve built a link, if you have a pulse – is that you should naturally vary your anchor text.

I hate this, and I hate it because it’s unhelpful and ambiguous. It’s like telling a fat person “eat healthy”. Yes, fat person, eat healthy, but if you only eat carrots and lettuce, you probably couldn’t lift a fly, and your diet will be so deficient in other areas you’ll likely develop fatigue, smug SEO consulting skills, or some other insidious combination of those and other negative characteristics.

My point is that naturally varying your anchor text isn’t enough. So, I’m here to help – first, to say why everyone says “naturally vary your anchor text”, and that’s it, and how we can get past that – helpfully – without ever saying “combinatorially, 93.99485% of your anchor text combinations should be your primary keyword”.

If All Things Were Equal..

We never, ever get past “naturally vary your anchor text”, mostly because SEOs are afraid. They’re afraid to say that 60% of the time you should say “Peas”, 30% “Good Peas”, and 10% “Green Peas”.

The reason for that is because anchor text variation, more than any other factor in SEO, is practically impossible to test – all of us, the SEOs, only have personal inferences into what moves the needle and what doesn’t in this regard. And also, it’s hard to hold that one variable static while accurately – or even with 51% accuracy – holding everything else in place. We know more indications of an anchor text mean a higher probability of ranking for a given keyword, but we also know that at a certain, high distribution diminishing returns occur, to the point where a penalty is possible.

Where do we draw the line? Where do rankings peak? At the same threshold, one must consider that even if rankings are optimized for a keyword at a given anchor text distribution, if a website was given a manual review, would it still pass the same sniff tests? Almost assuredly not – or we wouldn’t see constant manual tweaks to the rankings – since a absence of this would mean that Google believes their algorithm is perfect.

So where do we go from here? This post isn’t to give you reiterate “naturally vary your anchor text” like other parenthetical, useless, incorrect SEO phrases like paid links are bad, or content is king. The point of this post is to be like Stephen King’s “On Writing” – which states that there is no science to the craft, but there is a path.

“Naturally vary your anchor text” says throw shit on the wall, hang around your keyword, and hopefully you’ll rank. Obviously, this is useless. There’s a better process. There’s unnaturally natural variance.

Using Excel to Divide Anchor Text Distribution

After I’ve created a given keyword focus based on a set of URLs, I draft up an Excel sheet. Along the top row, I paste the URL of each keyword that will be targeted in each column (and sometimes, the keywords cross URLs – and thus, columns).  Once I’ve done that, I give each column a color code based on how heavy I want the URL to be focused on in link building. For ease of use, I put URLs that are focused on more heavily in darker colors.

From there, I freeze the top pane by going into the “View” tab in Excel 2007, hitting the Freeze Pannes dropdown, and clicking “Freeze Top Row”. This may different depending on the age and type of Excel program you use.

Within each URL column are the appropriate keywords, which, most obviously, inform which keywords are focused upon the most. Below each column header, I begin filling in the keywords for each URL, based on whatever I deem appropriate for my “unnatural natural” variation. When a link goes live based on my specifications (or without), I modify and/or fill in the column cell of the anchor text that went live to yellow, to indicate that a link had been achieved for that anchor text.

Given this data, I can make deeper inferences into why certain keywords are moving, why others aren’t, and what my next anchor text and URL should be.

Although tools like Raven let you add links and observe tools like this, I prefer this as a process as it’s easy, and makes visualizing your focus and anchor text distribution a straightforward and simple process. When the cells fill up with yellow side by side, you’ll begin to notice when one URL hasn’t been hit enough, you’ve hit one anchor text too much, or what links and combinatory distribution seems most beneficial to your niche and rankings. When you’ve done this enough, you’ll begin to make your own, more accurate estimates about the right mix of anchor text for your keywords, content and vertical.

And most importantly, you’ll stop having coffee with business people while muttering penetrating sayings like “naturally vary your anchor text”.

And Then, Actual Suggestions

Although I’m not going to make the bold move that means suggesting an exact anchor text distribution that’s optimal to rankings, I am willing to offer certain anchor-text-specific details I’ve found cogent in my career, and moreso, are overall paths to best current (and future) anchor text practices.

  • Using Tildes (~) as a methodology to find the most semantically linked keywords to vary your anchor text with. Tildes, when placed before your keyword, return other keywords, in bold, that Google sees as extremely linked to the right-adjacent query. Sometimes, there won’t be any additional keywords Google seems to seem as extremely semantically linked, but it doesn’t mean they see any. Anyways, if you can determine other keywords that are strongly linked to your keyword focus, it seems likely that they would do the biggest benefit to your primary keyword when you vary your anchor text. In example, Google sees terms like airline tickets and flights as closely linked to the term [airfare].
  • Including domain/brand names in anchor text. Occasionally including the domain name in the anchor text is just one way Google might interpret your website as a brand. I’m not sure that a certain ratio indicates brand to Google, but it would make sense that a given number, at scale, would show that the site is a brand in a given vertical.
  • Indicating exact match domains are brands, and not keyword spam, by capitalizing anchor text. Brands are nouns. Keywords like “cheap flights” are not. I go more in depth on this on my post on how Google might devalue exact match domains, but my overall thought is that capitalization in anchor text, over a broad enough linkset, is a strong indicator of a brand rather over a spammy exact-match domain that manipulatively retrieves links with “cheap flights” into the domain “cheap-flights.info”. Shoes.com and CheapFlights.com are frequently capitalized, after all – because they are brands. If you look into the backlink profile of cheap-flights.info – I’m pretty certain they don’t share the same characteristics.
  • Varying emphasized keywords with other long tail versions of the same keywords. I would love to see a citation of a high value keyword that does not have other long tail versions of the same keyword – that can’t be hit on the same URL. Sometimes this takes some nice information architecture, but if you can’t find SOME long tail keyword to hit on one of the primary URLs you’re linkbuilding for, you’re doing the link building, keyword research, or information architecture wrong. Seriously.
  • If you hear “naturally vary your anchor text” too much, you’ll eventually start believing it. At times early in my career, I would start hitting my main keywords one out of six links, even though the all the variations were really close to the main iteration. Even though the other five were derivatives of the 1st, this ratio was too thin to make a real impact on competitive keywords. “Vary” does not mean go on a marathon jog away from your starting keyword. Run up the stairs to your room, come back, and use the keyword again.

Anchor text isn’t an exact science. Just kidding – anchor text IS an exact science, and that’s the problem. We’ll never know for sure where it’s going, or what the ideal variation is because of the hundreds of variables we can never truly control. With enough rationale, experience and “feel” that comes with a experience and time with keyword set with it’s own unique semantic closeness – you might just get a good feel for a near-ideal percentage to maximize rankings for a given keyword.

I wouldn’t ever come out and say it, though. These things will never stay stable, and they’ll never be provable. So just know that you should use one anchor text 100% of the time, and not 0% of the time, either. That much we know for sure. ;)

Previous post:

Next post: