Thanks for the links and the advice, Marcus.

I think after reading through the material I will meta noindex any search that has more than one search filter applied. So I'll index "blue" or "vintage" but not "vintage/blue" for instance. The most important top level search filters will become category pages, more or less. I'll try to tailor their content to reflect their importance. Thanks for your input!

We have a product catalog with several hundred similar products. Our list of products allows you apply filters to hone your search, so that in fact there are over 150,000 different individual searches you could come up with on this page. Some of these searches are relevant to our SEO strategy, but most are not.

Right now (for the most part) we save the state of each search with the fragment of the URL, or in other words in a way that isn't indexed by the search engines. The URL (without hashes) ranks very well in Google for our one main keyword. At the moment, Google doesn't recognize the variety of content possible on this page. An example is:

http://www.example.com/main-keyword.html#style=vintage&color=blue&season=spring

We're moving towards a more indexable URL structure and one that could potentially save the state of all 150,000 searches in a way that Google could read. An example would be:

http://www.example.com/main-keyword/vintage/blue/spring/

I worry, though, that giving so many options in our URL will confuse Google and make a lot of duplicate content. After all, we only have a few hundred products and inevitably many of the searches will look pretty similar. Also, I worry about losing ground on the main http://www.example.com/main-keyword.html page, when it's ranking so well at the moment.

So I guess the questions are:

• Is there such a think as having URLs be too specific? Should we noindex or set rel=canonical on the pages whose keywords are nested too deep?

• Will our main keyword's page suffer when it has to share all the inbound links with these other, more specific searches?