If your site gives the user substantially different content from what Google spiders, the risk is a penalty or ban.
If you serve a 404 for search traffic on one keyword, then you're saying the original url does not exist. Google will most likely discover that, and then the url will stop ranking for ANY keyword.
However, this kind of Google Search traffic can be a help in improving your content. If there is significant traffic coming on a not-too-relevant keyword, I'd develop a new page that is solidly on-topic, and get that page to start ranking. Link to the new "Cat page" from the currently ranking "Dog page", for instance.
Msg#: 3883085 posted 9:31 am on Apr 2, 2009 (gmt 0)
I see this from time to time. Google doesn't rank the actual content page, but instead ranks a page pointing at it.
In some cases it can be a straight 'lack of PageRank for the content page' problem, other times it has been as a result of a site having a Duplicate Content problem and the 'real' page has had multiple alternative URLs which Google has ignored in favour of the other page.
I know Tedster has made comments about this issue several times in the past. I can't remember what he said back then.
I definitely wrestle with this problem. In the past I had one major struggle that took almost a year to get working properly.
The fix seemed to be broadening the semantic content of the page that "should have been" ranking - in other words, I abandoned the narrow focus on the literal keywords and allowed the content to range more widely into synonyms and co-occurring terms.
That was worth the effort because the problem was for one of the five top-level category pages - it's probably not a solution for thousands of individual product pages.
This is one of those areas where Google needs to do better, IMO. It frustrates me for the sites I work with and it frustrates me as an ordinary user of search.
It seems to be a result of anchor text as an ON-PAGE factor. A similar problem existing with blogrolls, and Google recently launched a new fix for that - maybe they will eventually use similar logic for other types of sites, too. I can only hope.