Welcome to WebmasterWorld Guest from 22.214.171.124
Forum Moderators: open
Results seem based on linkage and content
I disagree that the results are based on content. In my opinion, results are just detected from the linking structure in a way "pages that are linking to that page also link to the following pages". (Similar to Amazon: Customers who bought titles by **** also bought titles by this author.) Of course, there is an algorithm to filter out the most important related sites. But I didn't see any hint that Google is considering the topic of pages (for the "similar pages" results as well as the ranking algorithm).
In the local ODP category there are many sites listed but only 2 of them are seen as similar pages. Both might be considered theme related (one beeing a software developer, the other an internet consultant). One site is only listed in the local dmoz category, the sofware site is also listed in a software category in dmoz.
It is not necessarily the content that let Google assume they are similar but at least both have been seen belonging to a similar theme which must have been triggered by content.
However, looking at the results for a particular page, I noticed one page about "widgets".
No links to or from pages about widgets, but widgets are mentioned in the content.
As a Topic-Sensitiveist, I believe content will have an increasing bearing on things ;-)
In the local ODP category there are many sites listed but only 2 of them are seen as similar pages. ...
... but at least both have been seen belonging to a similar theme which must have been triggered by content
This doesn't show that Google is considering the theme. As already said, there is probably a filter to determine the most important results, but this is not necessarily done by triggering the content. There might be other reasons that only 2 pages are considered as similar, e.g. the other links are new or similar pages is PR dependent (as the link command).
Of course, I cannot solve the problem for this special case which I haven't seen.
as with most things Google, it's difficult to pin-down absolutely ;-)
No links to or from pages about widgets
It isn't necessary to have links to other pages/sites. You just need pages which link to both pages as (for example) the ODP.
By the way, even if you look at Amazon you will often find links to books which have the same theme (although they were just detected from people who bought books and not from the content of these books).
No position for dmoz category keywords now but I doubt they ever had a good ranking for those keywords. Just because a site is listed in a dmoz category for a city doesn't mean it should be listed for a search on that city. As for the software category it is also a very braod top category (software/industry-specific) where the site should never be found if somebody is looking for a keyword like "software industry specific".
"There might be other reasons that only 2 pages are considered as similar, e.g. the other links are new or similar pages is PR dependent (as the link command)."
There are other sites in the city cagegory with much higher PR. Also most of the sites in the city category are there for quite some time (years).
I stickied you details about the case I mentioned.
Looking deeper I found that related:www.siteA.com shows siteB.com. But related:www.siteB.com doesn't show siteA.com. SiteA.com is that is only in the city category of dmoz and SiteB.com is the software company.
I have examined several of my sites and normally the 'related pages' aren't related apart from the fact that a page exists which links to both. Moreover, I have also found an swf file which was considered as related. Of course, Google can find links inside Flash files but isn't reading the text.
Your mind seems set against the idea of content being a factor ...
My experience is against it.