How does google determine where it gets it's snippets from?

Sometimes it's autogenerated, and some times it's from another source.

[Mike12345]

How does google determine where it takes its snippet from in the desc. in the SERPS?

Is there an answer to this question, or do we just not know, or am i missing the answer to the question already posted somewhere here. Which is more than likely!

Any thoughts?

[annej]

If the page is in DMOZ it seems to use their description for the page. If the search term has a word or two in it that are also in your meta tag description Google seems to use that. Otherwise they take the snippet from something with the search word or term in it and then splats some other stuff in. It's such a pain. I wish they would just take the meta description all the time.

I'm sure someone will come along with a more educated analysis of how it all works.

Anne

[JudgeJeffries]

They seem to like the H1 tag

[NotePad]

i kind of like the snippets, from the searchers point of view, it shows you the context the keyword was found in, sometimes it's very usefull.

[progen]

splats some other stuff in

That sounds right to me. Sometimes it takes the first few words on the page and sometimes it puts together multiple occurances with '...'.

[yetanotheruser]

Mike,

Personally when we spider a page, all the text content is bundled together (ie, remove all tags and link content, special characters) so you end up with one long piece of text which is kept for this purpose.. Then for each search result to be displayed we look in this text for each search term.. personally I look for:

'a space' -> 'up to 40 characters' -> 'the keyword' -> 'up to 30 characters' -> 'a space'

If the keyword doesn't exist (ie, if it's just in the page title) you just get the first 100 characters or so..

I haven't perfected this by any means, it was just a quick regex solution, and I haven't got round to dealing with 'phrases' yet.. but it's very quick and simple and produces decent enough results that appear pretty similar to the same search on G.

IMHO I think there won't be anything particularily special to what Google (or all the se's) do appart from regexing their caches of the text content.. (Though I guess it might help us glean some insight into what text content they actually count in the algo, in our case it bears no relation at all ;) )

HTH's :) J.

[edit_g]

It can take anything it likes it seems - even things in <noscript> tags.

[Marcia]

I've seen a certain number of characters pulled right from the description meta, followed by the contents of an alt attribute used as the snippet.

Not sure how many characters in the meta description are used total except where I've actually counted, but I'll be tacking a bit on to one to see if it gets picked up in the snippet to try to get a little more accurate picture of what's a good length to use for the meta description.