When was the "similar pages" option introduced?

...and more importantly, why?

[glengara]

Every now and then I take a look at "similar pages" to try to discern a pattern, or work out the purpose of that search option.
The results seem based on both incoming and outgoing links together with a large dose of " je ne sais quois", and rarely bear any actual similarity to the page.
So can any Google historian remember when this search option was introduced?

[takagi]

It was first called 'GoogleScout'. If you search on that word, you can find out more about the history if this function. Like this press release [google.com] of Google on Sept. 21, 1999.

[plasma]

In fact it shows pages that have backlinks from the same page as the one you found in the SERPs.

e.g.

a->b
a->c

c is similar to b

IMHO this function is totally useless. It never brought up any 'similar' result.

[glengara]

I suppose that's what irritates me Plasma, it's unlike G to introduce and maintain for so long such an apparently useless option.
The only interesting thing I use it for is to get an idea of a pages link "theme".
I suspect a cunning plan somewhere ;-)

[vitaplease]

It tends to work well for established sites - but sometimes its not much better than just checking the relevant DMOZ cat.

If I do a similar pages to Google.com I get all the other search engines, but I also get some adult listings...

[glengara]

Thanks for that link Takagi, seems to have been introduced at the formal opening.

*Google delivers highly targeted and relevant results, but beyond that we offer GoogleScout, a revolutionary concept for locating groups of information sources on the World Wide Web. GoogleScout helps user surf the web smarter, faster, and easier.*

They seemed quite proud of it at the time, maybe the quality of the normal SERPs also took them by surprise, and made the need for Googlescout redundant?

*A GoogleScout link is created from the collective wisdom of the web. It delivers suggested websites that are based on an objective and fully automated analysis of web pages and overall link patterns of the web.*

Wouldn't this statement imply they expected the majority of links to be relevant?

[dirkz]

a->b
a->c

c is similar to b

In an ideal world, where links are just for visitors to get more information to site related stuff, this could work.

But unfortunately ... :)

What do you mean by

quality of the normal SERPs

?

[robho]

similar pages

This feature does work, but is mis-named. It seems to return "similar websites". In other words, it looks at the total theme of the website, and returns the home page of other sites with a similar theme. The content of the page is ignored.

For example a page on a global travel review site, mentioning nice hotels in widgetland, will for "similar pages" give the home page of other travel review sites. Even if they have nothing on widgetland.

It also seems to take into account common ownership of sites (or maybe just crosslinks).