When you say "the US datacenters" do you mean when you search using google.com - or are there specific IP addresses you are querying directly, by the number?

To my knowledge, the different data centers return "raw" results for many locations, and then some extra step comes in - something we do not yet undertand specifically. Something like a last minute filter.

All I can say is FINALLY! Our sector had been dominated by .co.uk and .au sites for ages - absolutely useless to US and Canada users, as none of the companies sell in these countries.

The changes I'm seeing are all .co.uk and .au sites (previously 6 out of 10 were of these extensions), are gone in google.com, but show in google.co.uk.

This occured this morning - Jan 29th.

Good job Google! :-)

There is a lot of odd stuff going on at various data centers. see this thread: [webmasterworld.com...]

I have several VPN channels with several IP addresses in differet countries. I check my ranking only on google.com from different IP adresses located arround the world.
For example my site ranks #6 on google.com for the UK IP addresses and it ranks #56 on google.com for the USA IP addresses for the same search terms.
My site has a .com domain name.

[edited by: Smark at 7:44 pm (utc) on Jan. 29, 2008]

I have a site with a .it domain. On every datacenter our main keyword phrase is ranked either #4 or #5 (datacenter IP search) but when I search on google.com we are always between #8-12. The server is USA based and I am searching from the USA.

For the singular version of the same keyword phrase we are #1 on all datacenter including a search on google.com. By the way the singular version of the phrase is much less competitive than the plural.

Does anyone know if there is some type of filter google applies to non-.com domains on the google.com results for competitive keywords?

I've never seen evidence that the domain extension plays a part in the relevance algo. Google spokespeople have said it doesn't. However, what exactly happens between the datacenter/IP and google.com results is still quite mysterious.

I've heard the same thing - that the domain extension doesn't matter - but it's hard to believe a google.ca search would be so different than a google.com search especially since there are no local .CA results for this phrase.. or any .ca sites for that matter that appear in the first 10 pages of the results.

Baffling.. and very frustrating.

"..that the domain extension doesn't matter ..."

IMO that may apply more to TLDs rather than ccTLDs, particularly non-Anglophone ones...

I've seen SERP differences for at least one of my sites on .com versus .ca versus .com.au.

I've contemplated making a specific country target in Webmaster Tools, i.e., to America, because most of the traffic is from the U.S., thinking it could help the .com rankings.

But I really have no idea if it would help google.com results or diminish the alternative tld SERP results.

I did notice one domain that was first used in another country, and got IBLs from other sites also from that country gets decent link juice from Google for that specific nation's Google results.

Not enough data to make that into a theory, but it wouldn't be completely irrational.

p/g

I've been looking at regional links/national Gs recently, too early in testing to announce anything "officially", but as of now I'd certainly look for local links when targeting a specific country.

I did some research and this is what I found.
My IP adress is in the USA.
When I perform a search on google.com my site appears on 50+ positions.
Then I did a ping test of google.com and found that it has 64.233.161.147 IP adress.
Then I searched directly [64.233.161.147...] - my site is in top 10!
Then again [google.com...] - my site is on 50+ position.
Different results on the same DC!

The results on the same data center differ depending on how it's accessed (by IP or by domain name).
Any ideas?

[edited by: Smark at 8:40 pm (utc) on Jan. 31, 2008]

Then I searched directly [64.233.161.147...] - my site is in top 10!
Then again [google.com...] - my site is on 50+ position.

This is exactly what is happening to me on my most important keyword. But it has nothing to do with what country. All the numbered data centers show the page near the top but Google.com shows it way down.

I've tried and tried to figure out what is causing this. It's like a final filter is put on it then it goes to google.com. Only the single word seems to be penalized. If it's combined with another word the phrase ranks just fine.

I have the same problems,
(see [webmasterworld.com...] )

To me, it doesnt look like it's related to discrepancies between datacenters; it looks more like a region filter intentionnaly apply by Google.

[edited by: tedster at 10:25 pm (utc) on Feb. 1, 2008]
[edit reason] link repair [/edit]

Does anybody with this problem have multi-lingual pages? If so, are the different languages in sub-directories (i.e. domain.tld/french/) or are they in subdomains (i.e. french.domain.tld/)?

I have a domain with .nl extension that is affected, but in a strange way. One single keyword query where my site has been around #10 for 5+ years is completely missing in the US Google SERPs, but that same keyword with any other niche relevant keyword is at its normal position in the SERPs or even better (many times #1 to #5)

Other high traffic single keyword queries for this domain are unaffected. Their position is the same or even better than before. One high traffic keyword rose in the last few weeks from #4 to #2 in the US SERPs (seen via a US IP based proxy) whereas it is #4 on the international english language www.google.com when looking from a NL IP address. So this specific .nl URL performs better in the US than it does in the Netherlands.

On average my US traffic to the .nl domain is slightly up in the last few weeks, rather than down.

What may help is that my the webserver itself is located in the US, so IP wise it is a US webserver. Furthermore I have strong incomming links from US based servers many from trusted neighbourhoods.

My first conclusion sofar:
It is not per se the domain extension that is targetted by the algorithm change, but it could be a combination of the webserver IP and the relative strength of links coming from US and non-US servers.

Geo targeting web servers is a pointless exercise as it's easy to buy server space anywhere you want. Still you'll only find one best deal, so in the end you'll end up with a single geo location. Kinda hard to address a global audience through that.

Anyway Microsoft was doing this for a long while and they "can't fix the problem", since they're now buying Yahoo!, that'll get infected soon too and now good old Goog in the same bad waters. A pity.

I've never seen evidence that the domain extension plays a part in the relevance algo

Who said anything about domain extensions (TLDs)?

This is apparently based on geo-location of your IP address. So, (a guess at) where your server is physically located.

I can see some problems with this, though. Although IP geo-locations is generally accurate to the country level (it's famously inaccurate when attempting to go to the city level), it isn't flawless to the country level, either.

For example, all Google IPs will geo-locate to the U.S. no matter where they are in the world, if only using the allocation tables.

Same goes for companies such as BT, many of the Asian backbone carriers, etc. and I'd imagine this may extend to some hosting operations.

Do traceroutes, and you will see that often the U.S. endpoints of overseas carriers geolocate to that country, not the U.S. You can find loads and loads of non-U.S. IPs that geolocate offshore but really (obvious from ping times and the rest of the route) are in N.Y. Florida, or San Jose.

Now, imagine what happens when a backbone provider also has hosting operations. I suspect in many cases they didn't think of the implications of their IP address use. So, an English company with a hosting operation in the U.S. might geolocate to England, or a U.S. company with a hosting operation in Germany might geolocate to the U.S.!

Geolocation can be enhanced with proprietary information (there are databases available, for example, that map-out the IPs of major end-user ISP's POPs - say Cox or Comcast (often you can figure it out from the domain names, they often contain abbreviated city names) and guesses based on route times, and I'd hope Google is doing this, but in any case it's not going to be perfect, and some sites will be miss-located.

Since our servers are USA based and we have this problem I doubt very much if this is a geo location problem.

Since our servers are USA based and we have this problem I doubt very much if this is a geo location problem.

Same to me.
.com domain and server in the US.
The company is outside the US.
This filter also hit some companies in my sectror. They are mostly non-US.
They have .com domains and other domains.
But not all non-US companies were hit in my sector.

I don't know how G can distinguish a non-US business from a US business.
May they manually check top websites in some product sectors?

I used to be hosted outside the US with a company that was a regional office of a large US hosting company. So while my hosting was not in the US, I was for all intents and purposes on a US IP address. Geo targettings pretty good, but it's not perfect.

I just checked a site of mine. I've got a regional site with many regional links, and hosted in the region, that ranks in the regional Google. It's rankings have not changed in Google.com - it still ranks pretty good.

I think that people are being too simple when thinking about how Google might understand the geo-target or location of a site. The geo-location of the IP or hosting might play a part, but it is not the only tool they have in their hands.

I am guessing they are looking at a combination of factors to determine geo-location. We forget the sheer amount of information avaliable. They could technically tell the geo-location of a user if:
A) they have the toolbar installed
B) (possibly) if they are using FireFox
C) they go to a site that has AdSense directly after visiting your site
D) they go to a site that uses Analytics directly after visiting your site
E) they go to a Google property directly after visiting your site
F) if your site uses AdSense
G) if your site uses Analytics
H) if your site uses AdWords

Think how many users that covers. They can get a decent sample and tell from your user's locations where your site is really targeted to.

Not to mention the geo-location of links into the site.

Google never does anything as simply as one factor. Why should this be any different?

I've recently run into some mentions (from Googlers) of Google's potential to use historical records of user satisfaction with any given SERP entry, and to do that by IP block.

In other words, Google already has a history of their search end-users by their IP address. If a given url seems to satisfy users from a certain location, then that factor might provide a boost for that result, but only for that geo-location. And of course, if users from those IP blocks don't seem satisfied on average, then the url's rankings might drop for that geo-location.

It's one more factor that could be in the mix.

I've never seen evidence that the domain extension plays a part in the relevance algo. Google spokespeople have said it doesn't. However, what exactly happens between the datacenter/IP and google.com results is still quite mysterious.

I have. For more than 1 year I had a graphic image from one of my .com sites displaying at the top of the regular SERPS for a 2 keyword phrase. It resulted in some very good traffic.

Then a user of a uni .edu site .. not the site per se,but the daughter of a professor .. illegally downloaded that image. Google then substituted their URL .. same image. Traffic drop was about 50%.

All my German blogs on Blogger (I always use my US Blogger login/account as I run 50% English and 50% German sites) suck in German DC's.

Is there any way to improve? G doesn't seem to care.

These changes seem to have been undone. Results are back to what they were before. Roll back? Anyone else seeing this?

I am not seeing a rollback, and have experienced exactly what others have - with the timing being identical as well.

If I use a proxy browser (located in the US) and search for widgets we come up in x position, and if I search normally we're in totally different spots. This wasn't the case 3-4 days ago, both proxy and local serps matched, or close to it. Plus the .co.uk and .au sites that were appearing in the normal searches were gone, now they are all back. So what I'm seeing in our sector would be a rollback to what was there a week and a half ago.

Very bizarre....

Matter of interest, did anyone notice a boost for .us domains?

While running a few checks recently on my site, I stumbled upon information which concerned me. It appeared that google believed that my site and company was based in Afghanistan. How did this come about? It was the ignorance of someone at my host company who incorrectly assumed that my country code was AF. This data was submitted to ARIN for the range of IP's that we own for our dedicated server, which is based in a large US city. When I logged into the server remotely and opened google in a browser, it defaulted to google's Afghanistan site (com.af). I did a search for keyword-keyword and it returned a large list of results. Then I searced for "keyword-keyword af", and only two results appeared, with my site at position 1.

Now we have absolutely no connection with Afghanistan and our company is not even located on the same continent. So I am just throwing this info into this debate on geotargeting as I believe google is factoring in the information from ARIN. This may well have a massive and potentially disastrous impact on a site's positions.

Naturally, I requested the host company to rectify the data asap, which they did. So far, it has not lead to a change in regaining lost ground on google. (one week later)

Does anyone else think that this is possible or probable?

Arin details for IP's can can be checked at [ws.arin.net...]