MHes - 8:07 am on Aug 13, 2004 (gmt 0)

Jesse_Smith

I'm sorry, you're tests are flawed. There are so many reasons why, that it is difficult to know where to start and as I do, I instantly lose the will to live.

Apart from the obvious constant algo tweeks that make your conclusions today, worthless tomorrow, the fact is that Google and the web is primarily based on links... a small matter you sweep aside.

What about these factors concerning your test.

1) Relationship of broad match words. Hiltop will not trigger if a nonsense word is used, and likewise other filters possibly applied to h1 etc. will probably not be activated.
2) Relationship of number of occurances of keyword, used in h1 etc. to the number of occurances in normal text.
3) Position of keyword within normal text throughout the page.
4) Occurance of keyword in title tag may trigger a whole new set of rules. Likewise, different combinations may trigger different rules.
5) Occurance of keyword in h1 but not in any anchor text out.
6) Overall word count and keyword density
7) Relationship between several pages on a domain all heavily focused on one keyword. On one page the h1 may be being ignored because several other pages use it, Google may have a duplicate content issue with your pages and be ignoring many of your test factors on one page and not another.
8) Word proximity within the tags, in combination with use of bold, word count and keyword postion throughout the document.

etc. etc.

Comments like "So it's official, Google doesn't look at meta tags if it's not a framed page, er didn't we allready know that?! " are just not true. You can never be so sure. Do they use them if the page is framed? I don't think so, but who knows. Do they use the keyword tag if the keyword is also in bold and h1 and links in? Do they use them if the word is in a noframes tag? They might... despite current theories to the contrary. The variations are too infinite to be so sure, nothing is official, we just have small snippets of observation for a particular example, which may not apply elsewhere with different circumstances.

Google's overall algo is probably amazingly simple, but within that simplicity is immense subtlety and complexity generated by factors in combination. You would have to do 10,000 test pages to even start making definitive conclusions, but even then, they would be based on test pages within the same domain and for a nonsense word, a huge factor that could isolate the conclusions to only that scenario.

The google algo is trying to find quality content which is relevant for a search word/phrase from a range of domains. On page factors will be related to links in from other sites and internally. They cannot be isolated.