Robert_Charlton - 10:30 pm on Jan 11, 2012 (gmt 0)

I've spent some time checking out search targeting for a client's site, and noticed that changing core keywords in a query could have huge effects on the types of sites and the Google reading level profile returned.

In general, overall vocabulary, sentence length, depth of information, etc, were apparent influencers.

Eg, a query on the level of, say, magazine article vocabulary brought up sites with this reading level profile....
- basic - 32
- intermediate - 40
- advanced - 27

But, searching for synonymous terms, using 2-syllable vernacular term for core keyword, this distribution came up...
- basic - 79
- intermediate - 19
- advanced - <1

My guess is that the search queries bring up sites intended (consciously or unconsciously) for the audience that would use them, and the entire vocabulary mix of the page and the intention of the site relates to the query. So this becomes as much a demographic targeting question as a search targeting question.

We discussed something similar in the early days of Panda with regard to health query results....

- a search for [diseasename symptoms] or [diseasename treatment] would give you high quality clinic and public health sites...

- whereas, a search for [how can I tell if I have diseasename] or [what to do if I have diseasename] brought up a great many content farms.

Google may ultimately calibrate reading level preferences further via personalization, and will probably do everything possible to determine how these preferences affect other preferences. I think it's much more complex than simply educational level, though. Slang terms, for example, often skew Google's tool toward "basic" level, but knowledge of slang can be highly specialized knowledge.