Anchor Text: How Many Words Work? Options

[kevs]

I asked a similar question recently about title tags.
For anchor text, how many words will Google acknowledge?
Some peole put 50 words. Is there any studies on it? thanks.

[Ruud]

#1

A good way to test this is to setup a couple of pages and link to each with ever longer anchor text, making sure that the words used in the anchor do not exist on the page you link to.

After a couple of days/weeks check if the pages are indexed. Then do a search for [site:example.com keyword] and see if the linked page comes up.

See Shaun's 2008 test of how many words or characters does Google count in a link?

#2

MP3's often have tags: artist, title, genre, etc.

How useful these tags are depends on you: if you label all your songs "music" only, you don't have a lot of use for tags. If all are labeled "music", because that is what they contain, and some of those are labeled "rock" while others are labeled "trance" ... then you have some valuable use.

Anchor text -> keywords -> tags -> to make retrieval easier, faster, efficient, relevant. More tags isn't always better.

[mrgoodfox]

I thought its good practice to keep the anchor text short and put the extra information in Title tag of the link.

I usually limit my anchor text length to 3 or 4 words at most.

[Ron Carnell]

See Shaun's 2008 test of how many words or characters does Google count in a link?

I would take that article with a very large grain of salt and encourage everyone to try their OWN tests. How a test is conducted will often determine the results and, frankly, there are enough questionable points in that article to cast the entire thing into doubt.

[EGOL]

Use anchor text that makes sense to the visitor.

[iamlost]

I'm going to extend this topic somewhat as 'how many characters count in anchor text' is about as important as asking how many angels can dance on the head of a pin or how many times a keyword should be repeated in page content - not very.

Of more interest is:
* extended anchor text or some amount of text surrounding anchor text - note that this can be divided into text prior and text subsequent, that they can be of variant size, and that they can be weighted similarly/differently.

* that both anchor text and extended anchor text can be used to describe and/or weight pages for terms not actually in content.

As character maximums are wholely arbitrary defining some current limit may be interesting but not especially important. What is important is developing a cohesive anchor and extended anchor text offering. This is fairly straightforward with internal links somewhat more problematic with IBLs.

[Ron Carnell]

Those are, indeed, extremely interesting questions, iamlost. Please, by all means, devise a test or perhaps a series of tests so we can explore the potential. Or, at the very least, get Google to cough up the specs for us.

Some things I know to be true, some I know to be false, most simply reside under an umbrella of speculation.

[A.N.Onym]

On a recent UX conference, it was said that a really well scented link has 12 *words*. Something to keep in mind regardless of any SE effect.

[Michael_Martinez]

It's not the amount of anchor text so much as the page providing the link and the page being linked to that matters the most.

However, the fact that test relies on nonsense words means certain other considerations were filtered out, so the test results are unnatural. That is, they're not very extensible to real-life search optimization.

Like Ron suggested, people should do their own testing -- particularly because each of us has our own unique set of resources and our test results may not always match other people's test results. However, any time someone creates a site specifically for a test, or makes up a nonsense word or expression to simplify their tracking, they leave the realm of natural search optimization and invest their efforts in unnatural search optimization.

The applicability of what you learn from unnatural SEO to what you can want to do with natural SEO is highly questionable.

[AbleReach]

The applicability of what you learn from unnatural SEO to what you can want to do with natural SEO is highly questionable.

[Ron Carnell]

Until Skynet makes its appearance, a word is a word is word. To a computer program there's no great difference between the words "natural" and "Quenya." Oh, wait, Quenya only used to be a nonsense word; now it's all over the Internet.

"Natural SEO" is an oxymoron.

[3rdeye5]

To a computer program there's no great difference between the words "natural" and "Quenya."

I believe that Google does apply linguistic analysis. At the very least there is synonym data and frequency of occurence on the web. That makes nonsense words a whole different category than real words.


Ewald

[glyn]

Make content relevant an on topic, organize the website properly and the anchor text CAN ONLY BE RELEVANT by definition.

[Ron Carnell]

Google does, indeed, process words to a point far beyond anything conceived of just ten years ago. Mostly, I think, it looks at a lot of lists. That does not, however, mean that Google knows how to read and process information like a human being can. It's tempting to anthropomorphize a computer program, but it's a temptation we should resist.

What is a nonsense word?

Unless a programmer can unequivocally define what is meant by "nonsense word," they won't be able to incorporate the process into a computer algorithm. Most names, for example, aren't in any dictionary. And it was only a few short years ago that "blog" would have qualified as utter nonsense to most of us. Today, a word can go from nonsensical to topical in a heartbeat.

As human beings, we could probably agree that a nonsense word is one that has no meaning assigned to it. But a search engine doesn't assign meaning to ANY words, because it doesn't really understand meanings. At best, it tries to establish relationships. That is perhaps a subtle, but very real distinction.

Here's the bottom line, though.

You can take a random series of letters, void of meaning and probably unpronounceable, stick them on the end of a very long Title tag, and when the page is eventually indexed, you can then perform a search on that random series of letters. If your page is returned in the SERPs (and it will be), that gives you some important information about the way the search engine indexes page titles. You can, if you wish, conclude the SE only indexes long titles if they contain a random series of letters, but that conclusion is going to be of very limited value. You have to be willing to move (carefully) from the specific to the general if your experiments are to be useful.

Over the years, things like off-page factors and duplicate content filters have made it increasingly difficult to devise meaningful SEO tests. Being able to search for a unique series of letters, i.e., a nonsense word, is still a valid tool, and I believe it will be until programs learn to actually read. It is essentially nothing more than a filter, one that can help eliminate a few billion documents you don't want to see. That's, after all, what a search engines is designed to do.

[3rdeye5]

Most names, for example, aren't in any dictionary.

Not in dictionaries per se, but in phone books they are. Information about people's names, brand names, organisations' names are abundantly available in various kinds of online lists. I'm pretty sure that Google can and does differentiate between dictionary words, misspellings, names, screen names and words without meaning. Google probably uses lists categorizing what known words are.

It is pretty easy to distinguish a nonsense word from one with meaning: enter it into the Google search box and see how often it comes up. If it comes up a low number of times, it is likely nonsense. If it is used a low number of times AND it is used in obvious spots for screen names, it is not. If it is a known misspelling, it is not.


Ewald

[Ron Carnell]

LOL. Don't you think that's a little circular, Ewald? Google will know if it's a nonsense word based on the SERPs? And, uh, the SERPs are affected by whether it's a nonsense word?

If you're going to incorporate nonsense words into a computer algorithm, you can't use that same algorithm to define what they are. It's a chicken and egg thing.

I should also mention that when you're trying to define something for a program, like nonsense words, you should be careful not to introduce something else you'll need to define with equal precision. Like screen names?

Again, the bottom line is that they work.

[bwelford]

Did you realize there are 1.2 million SERP references in Google to gobbledygook?

Just thought I'd mention it.

[iamlost]

It's not the amount of anchor text so much as the page providing the link and the page being linked to that matters the most.

There is gold in that sentence. In more ways than one.

However, any time someone creates a site specifically for a test, or makes up a nonsense word or expression to simplify their tracking, they leave the realm of natural search optimization and invest their efforts in unnatural search optimization.

It is quite possible to use an existing established site and both new pages and existing pages to test real applicable terms within and around anchor text with both internal links and IBLs. Or using nonsense words. And they may or may not prove similar behaviour. Which could be interesting in itself.

To be successful a test must prove, or not, a prior hypothesis in such a manner as to be (1) replicable and (2) within set boundaries identifiable by recognised consistent outlier results. Very few 'SEO' tests that I have seen published meet these requirements.

That many/most 'SEOes' chase unnatural search optimization shortcuts is, quite frankly, a personal delight. Indeed the industry thrives upon mis-connecting disparate dots and believing six impossible things before breakfast.

[3rdeye5]

LOL. Don't you think that's a little circular, Ewald?

No, I don't.
Let's do an experiment, as the proof is in the pudding. I'm putting here the nonsense word: "ffoklop".
Let's see whether Google will confirm my hypothesis, when this webpage is indexed.
We can compare the results with those for my screen name "3rdeye5." (And no, that guy on Tripod is not me.)

Did you realize there are 1.2 million SERP references in Google to gobbledygook?

So, "gobbledygook" means something!
It is in the dictionary, btw.


Ewald

[Michael_Martinez]

You can take a random series of letters, void of meaning and probably unpronounceable, stick them on the end of a very long Title tag, and when the page is eventually indexed, you can then perform a search on that random series of letters. If your page is returned in the SERPs (and it will be), that gives you some important information about the way the search engine indexes page titles.

That is utter nonsense, Ron.

The more unique the expression the fewer factors a search engine has to evaluate in order to determine a ranking.

Natural SEO deals only with real query spaces that people are actually using (both as searchers and as publishers).

You cannot reverse engineer a search engine's algorithms. You can try to identify trends in the data the search engine provides, but the more removed that data is from what natural, everyday searchers are looking for, the more contrived, artificial, and irrelevant your data becomes.

SEOs are generally terrible at figuring out what search engines are actually doing. It takes hundreds of people throwing out ideas on dozens of blogs and forums before even the simplest principle begins to reveal itself.

These cheap little stunt tests where people create new sites and throw nonsense words on them (which any competent linguistic programmer could identify comparing them to a database indexing the vetted corpus) only act like bells and whistles that pump up everyone's adrenaline. You learn nothing useful from this kind of faux science.

When people start searching on ixbywizzymoog then other people will start writing content for it and eventually you'll have a well-populated query space where all the silly little SEO tests fail to work.