One of the projects I'm working on at the moment involves offering users related content once they've consumed whatever content is on the page. It's a bit like the Amazon-esque "users who liked this also liked" idea, though in fact recommendations will be based on semantic relationships between content rather than on other users' behaviour. You can assume that links to each piece of related content would constitute a thumbnail and a few words of text describing what that content is about.
This is a web-based product for a very broad user base who will come to the site with varying levels of expertise. My concern is that knowledgeable users who know what they want will have no trouble scanning through 10, maybe 20 related content items to find the one they want (i.e. the items that best match what they are interested in), but that someone who isn't very familiar with the domain, and maybe isn't sure what content they should engage with next, might feel quite overwhelmed by 10 or 20 choices and might be better served by maybe 3 to 5 suggestions.
In the interests of serving less expert users well, I'd generally err on the side of including fewer choices, not least because users who are more familiar with the content are probably in a better position to use the Search function if they don't see the related content they want in the initial recommendations. However, I can't assume a very high level of tech-literacy — even users who are more knowledgeable about content on the site may be quite unlikely to think of or use site search, and may in fact just go back to Google and try again with different search terms.
Has anyone encountered this problem before and found a solution that was satisfactory to the vast majority of users? Please don't break out the old 7 +/- 2 here, as that really doesn't apply (for reasons summarised quite well in this comment). I'm talking about visual and semantic overwhelm when given multiple options.