Should rel=canonical point to page one of a set and still have rel=prev/next on the page as well?

Question

I've got a page which displays search results and I've added rel=prev and rel=next meta tags to the page to cover the paging. However, someone is saying I should also add a rel=canonical tag to the page that points to the start page of the search results. So for example, say we're on mysite.com/search/xyz/page5 the canonical tag would point to mysite.com/search/xyz/

Is this wise? isn't the purpose of the prev and next tags to tell the SE that all these pages are of a set? But then putting the canonical tag in there is sort of telling the SE that the first page is unique and not part of this set?

http://www.youtube.com/watch?feature=player_embedded&v=njn8uXTWiGg#!&t=7m6s

According to that video from google, it sounds like we should not use the canonical link. But i just looked at ebay and they are using the canonical link instead of prev and next. Then again maybe they are showing a different page if you are bot instead of a user.

The problem we are having is that if we go into google webmaster tools it's showing all the paged results on our site as having duplicate content (duplicate title and description). Is that really a problem? Do i need to do something kind of stupid and just put the page # in each title and description?

Answer 1

You can use both, according to Google's documentation. Whether or not you should depends on the circumstances: if you've got other duplication problems, separate from the pagination problem (as per the example in the documentation), then you could set a canonical link element (CLE) for each page of the series – so the CLE manages duplication on each page, and the "prev" and "next" attributes deal with the pagination.

This said, it seems you're talking about internal search results: arguably, they shouldn't be indexed anyway. Google Webmaster Guidelines suggest preventing crawling of such pages with robots.txt. The problem with this method is that, if someone links to search results, they'll still get indexed (albeit with a "blocked by robots.txt" snippet). I recommend applying "noindex" (either HTML meta or X-Robots-Tag HTTP Header) to them instead.

Answer 2

You probably should not add a rel="canonical" element to your paginated pages, unless it is specifically required. The canonical link element is not required for pagination, it is used to resolve canonical URL issues. The canonical URL of a page in a series is probably not the first page of that series. Your rel="next" and rel="prev" elements already provide enough information for Google to work out the URL of the first page of a series. However, if you had a "view all" page (as suggested in the video) or you had additional URL parameters that are not part of the canonical URL (such as session IDs etc.) then go ahead and specify the canonical URL.

Google usually returns the first page of a series anyway in its results.

all the paged results on our site as having duplicate content (duplicate title and description).

Duplicate content is not the same as duplicate titles and descriptions. The content is presumably not duplicate, since it should be a different page of results. Including the page number in the title could well resolve the warning in GWT. This would also be beneficial to your users, since this is what they see in their tabs, history and bookmarks. Note that this warning in GWT is only advisory, it is not necessarily suggesting something is seriously broken and your site is being actively penalised for this. However, as mentioned, including the page number (or something unique) in the title of the page could benefit your users and if it benefits your users it can benefit your ranking.