ronburk - 6:06 pm on Mar 25, 2006 (gmt 0)

Basically one doesn't need to separate the words for the search engine to be able to identify them.

This myth is repeated periodically in various forums. The reason people tend to believe this myth is that they think that Google SERPs highlighting has something to do with how Google parses words for indexing.

In fact, Google SERPs highlighting consists of simple text substitution and is in no way indicative how how keywords are parsed for indexing.

Thus, for example (and I hope rather harmless specific searches/URLs may be forgiven to prove this very specific point), one can search for

site:www.mysql.com my

and the Google SERPs will highlight "my" when it appears in "www.mysql.com" in the SERPs listing. And then one can search for

site:www.mysql.com mys

and (because there happens to exist at least one page in that domain with the nonsense word "mys"), the Google SERPs will highlight "mys" when it appears in "www.mysql.com" in the listings.

If the naive belief that highlighting in the SERPs indicates how Google parses keywords were true, we should then be able to Google for this term:

inurl:mys "initiative to migrate to MySQL and the open source LAMP stack"

and find that among the hits is the mysql.com page containing this lengthy exact phrase. In fact, Google finds no hits for this search. Finally, as a sanity check, we Google for:

inurl:mysql "initiative to migrate to MySQL and the open source LAMP stack"

and find that it does find the exact page in the mysql.com domain containing that lengthy phrase.

It is therefore easy to see that the fact that Google may highlight the string "mys" in the SERPs text containing "www.mysql.com" in no way means that it ever parsed or indexed "mysql" as though it contained a separate word called "mys". It did not. Likewise, it never parsed or indexed "mysql" as though it contained a separate word called "my", as can be verified by searching for:

inurl:my "initiative to migrate to MySQL and the open source LAMP stack"

The quoted assertion is, therefore, definitely not true. You absolutely cannot run words together in text, URLs, or anywhere else and expect Google to parse them as multiple words.

Stemming and other simple grammar transforms are another issue, of course. Many search engines are happy to see the word "homes" and index it as "home", "homes", or both.