When Paypal’s official Web site no longer ranked #1 in Google on a
search for “paypal,” it was obvious that Google had become more aggressive
in penalizing sites with “unnatural” backlink anchor text. Although the
high-profile Paypal example has since been rectified, thousands of
webmasters are suffering the consequences of not ranking for even their
official company name, let alone their top keywords. It is important for
search engine optimizers to understand both how anchor text penalties
are being applied and how LSI ensures that anchor text variance will not
dilute a link popularity building campaign.
Anchor Text Penalties
In the past year, webmasters have found that the aggressive link
popularity building tactics that work well in search engines such as Yahoo!
do not fare well in Google. Google has implemented several features to
filter out sites that appear to have an unnatural backlink structure;
one of these features seems to be specifically penalizing sites with
unnatural backlink anchor text.
It has always been an SEO best practice to use descriptive anchor text
in both external and internal links. But search engine optimizers have
often focused on a single keyword phrase when choosing anchor text,
especially if their topic has one keyword that receives vastly more
traffic than any secondary keywords. Since good links are hard to come by,
they do not want to “waste” any of those backlinks with anchor text that
does not contain their main keyword.
The drawback to this approach is that it can be interpreted as
unnatural by a search engine. A site with organic, passively-obtained backlinks
will have a wide variety of backlink anchor text variations such as:
“official site title,” “keyword,” “keyword synonym,” “http://www.thesite.com”
and even “click here.” If the vast majority of a site’s backlink anchor
text is simply “keyword,” it is obvious to an algorithm that the link
popularity was not obtained organically.
Latent Semantic Indexing (LSI) Basics
Let’s now touch upon the myth I mentioned before, that if a backlink’s
anchor text does not contain your Web site’s main keyword, its power is
wasted. The concept of latent semantic indexing, which may be more
fully implemented by major search engines in the near future, will prove
this myth to be false.
Latent semantic indexing can help overcome the “vocabulary mismatch”
problem when a human uses a search engine. Individual words do not always
provide reliable evidence about the conceptual meaning of a document.
For instance, a Web page that is highly relevant to the term “laptop”
may never use the term “notebook,” however it is clear to a human being
that “notebook” is often used as a synonym for “laptop.”
While it is beyond the scope of this article to discuss the mathematics
behind LSI, its implications for search algorithms are simple. LSI can
use statistical techniques to create a semantic analysis for any given
query topic. In practice, this means that a page can be considered
relevant for a particular keyword, even if it does not contain that
keyword. For instance, a page that is considered relevant for “laptop” can
also be considered relevant for “notebook” even if it does not contain the
word “notebook,” if LSI determines that “notebook” is semantically
related to “laptop.”
The principle can be applied to backlinks as well. Backlinks with
anchor text that do not contain your Web site’s main keyword, but instead
contain a synonym or related word, may still be giving your site a bonus
for the main keyword.
Link Popularity Building Best Practice: Vary Your Anchor Text
The recent increase in penalties given to sites with unnatural backlink
anchor text, along with the possible implementation of LSI, should give
webmasters motivation to vary their backlink anchor text heavily.
Rather than seeking to only obtain links using their main keyword,
webmasters should include synonyms, variations and related words. Certainly no
single keyword variation should be used the majority of the time;
rather, the text of all links should vary widely, just as they would if the
links were obtained passively. This will ensure a site’s improvement in
the SERPs, without drawing a penalty flag.