fred9989 - 1:26 pm on May 30, 2012 (gmt 0)
I'm pretty sure that what I'm seeing for my own site is a phrase by phrase issue. Having said that, in my serps there are some whole sites that have suffered for all or at least many search terms.
I'm just hypothesising that there is something related to anchor text on and off site and keyword stuffing in certain areas (types of tags like image alts for example) of pages that cause the term by term effect. Anchor text in sitewide links like the big footer on every page with links to "key" pages on the site may also be involved in this.
Then there's a different ssue that affects a whole site that is caused by link buying, hidden back links, links from certain networks etc.
If you have both then Ker-Bam! you're gone.
I see an extremely high algorithmic sensitivity to the combination of -
1) keywords in anchor text
The specific words or phrases in anchor text as a % of the total links into a site now has to be lower than ever; above certain low limits - which are somehow linked to points 2 and 3 below - a site is pushed into what used to be called +950 penalty. And the algo seems to treat related meanings as identical, so several words in different anchors may not protect a site - an example of this would be the words stop, control, prevent, avoid)
2) keywords in meta title
if the title resembles the anchors, or the domain, it's down the hill by 10, 30 or more places
3) keywords in domain name
I begin to think this seems to be the worst crime of all in Google's eyes! In one of my niches, the results show almost no sites with anything like the key words anywhere in the URL, let alone the domain name, except for wikipedia.
I suspect the analysis of these 3 factors is done almost phrase by phrase or at least by very narrow area. Certainly it looks like all the EMDs that have the same keywords in the meta title are gone from my field, a fact at variance with some other commentators' conclusions (that EMDs have not been affected because of the need to keep things like Amazon near the top of the results, assuming amazon is a key word for amazon, if you see what I mean).
I have one site which I have not been optimising for much, with a domain name unlike the subject matter, and non-optimised text, bar the meta titles, which are relevant, which has burst out at the top of the search results.
In fact it astounds me how quickly the site has been indexed, the links picked up, and how Google knows when to return this site for - I think that's very smart...... As far as a piece of engineering about presenting results which are relevant and non optimised I must admit it's very clever. Unfortunately the rest of the sites around it are pretty cr*p.