Google related: Search

Does anybody know how this is determined?

[djgreg]

Currently I am very astonished which websites Google thinks are related to one of my domains.
There are three types of sites in the list:
1.) Sites of mine whích are hosted on the same server but have a completely different theme and which are not linked to each other.
2.) Sites which link to the site in question but are hosted on another server.
3.) sites which have absolutely nothing to do with me. Other server, no linking, even another language.

I can imagine that Google determines related: through the linking structure but why are there so many unrealted sites in the results?

[g1smd]

I found that a page of my site hadn't been cached, or re-indexed, for a very long time. The page showed up for searches where it should not. There was no longer any related content on the page for the search it appeared in, and the snippet also reflected very old content.

Several other people that I know also had similar problems with odd pages on their sites. I assumed that putting up a fresh link, from a high-ish PR page, to the affected pages might tempt Google to have another look at them. So, on one page of one my sites, I set up some links to these affected pages (about 9 or 10 links) on the other sites. Several months later I saw that most had been reindexed as expected (though several remain with one year old cache, and supplemental result).

On one of the searches that I did, to verify that the new content had been indexed, a friends site was the only result returned. I was curious as to what the "related sites" link might show. I clicked it and was surprised to see that it was simply a list of all of the sites that my site had linked to to try to get them reindexed.

Those sites are not related by topic or by location. They are related only by the fact that one page of my (also unrelated) site links to all of them.

So, the "related sites" are more likely a list of (some of) the "forward links" gleaned from one or more of YOUR backlinks.

[edited by: g1smd at 8:15 pm (utc) on Feb. 9, 2005]

[internetheaven]

All the "related" sites I see in relation to my sites are those that are linked to by the sites linking to me.

i.e. if ten sites link to me AND another specific site Google thinks that it must be related.

That is the only common thread I've seen, all the sites that Google says are related are linked to on pages that I am linked to from.

[djgreg]

yes that's what I thought, too. But there must be something more behind it.
To promote one of my domains I traded links with several other webmasters and I also set up links on other domains of my own to push PR.
In my understanding Google should show all sites which link to me from more than one page in the results of the realted: search.
Surprisingly it only shows MY domains, as it would want to say "Hey, we know that you promote your own domain by interlinking it with your other domains."

The domains are not on a single IP nor on the same server.

greg

[BillyS]

For me related: show sites that are in the same DMOZ directory (in addition to others).

[AjiNIMC]

I call it related site syndrome, google sometimes use this to catch three way linking pattern.

I can say it is a part of LSI,

Pagex of Site A has links to yousite.com and mysite.com
Pagey of Site B has links to yousite.com and mysite.com
Pagez of Site C has links to yousite.com and mysite.com
..... (it continues t times)

And take this case
Pagex of Site A has words n-dimensional and topology
Pagey of Site B has words n-dimensional and topology
Pagez of Site C has words n-dimensional and topology
.... (it continues t time)

A search for n-dimensional will therefore return a set of articles containing that phrase (the same result we would get with a regular search), but also articles that contain just the word topology. Now if Google says that n-dimension and topology are similar words , it can also conclude that yoursite.com and mysite.com are related.

Other Reasons
Check Whois for both the sites

Aji

[rfgdxm1]

>For me related: show sites that are in the same DMOZ directory (in addition to others).

Which in theory makes sense. At minimum, if the ODP editors are doing a good job, sites listed in the same cat as yours *should* be on the same, or at least a related, topic.

[pmkpmk]

We have a lot of documents in PDF format on our site, which were generated by Quark Xpress. A "related" search on our site shows www.quark.com as being related to us. There's only sites featured as related, who are remotely related (at best) to us. Fortunately though, not a single of our competitors is listed as related :-)

[djgreg]

AjiNIMC

Do you really think WHOIS data is used to find related sites (related sites as sites from the same owner)

I can't imagine that because it would be very complicated to compare & collect those data.

And what about the privacy of the data.
In germany for example you need to agree that you will not use whois data for anything than legal or administrative issues. I don't think that it would be possible for Google to use this data in any autmated way.

greg

[AjiNIMC]

Do you really think WHOIS data is used to find related sites (related sites as sites from the same owner)

I cant give you a complete proof but SEs can always use the whois to check relationship.

I can't imagine that because it would be very complicated to compare & collect those data.

No Not exactly, when people like us can program for it then anyone can.

And what about the privacy of the data.
In germany for example you need to agree that you will not use whois data for anything than legal or administrative issues. I don't think that it would be possible for Google to use this data in any autmated way.

They may not accept the fact that they are using it to track related site networks, but you never know. Its always advisable to take a note of this.

Aji