Indeed, earlier incarnations of LinkSpam would try to associate the spam link with paragraphs of relevant keywords, almost making the spam something akin to content; however, since Google's PageRank algorithm cares mostly about the LinkText, it's become sufficient to ShotgunSpam named links into public spaces, e.g. [named link].
The goal of LinkSpam is very simply to suck off PageRank from your site and give it to the spammer's site. SocialSoftware sites are great targets for LinkSpam because sites like wikis and WebLogs both naturally accumulate PageRank as well as often allow anyone to add hypertext (via editing the page or commenting). As a result, LinkSpam has been on a meteoric rise through 2004 and 2005, particularly after the LinkFarm?s were defeated in the Spam Wars of 2003.
LinkSpam has an intrinsic PowerLaw tension. On one hand, it's best to spam highly ranked sites in order to maximize the value of the spam. However, these sites are also heavily moderated and protected, making the effort often pointless if the spam is erased. VersionHistory (and archives) may defeat the benefit of moderation, even if NotIndexed or NoFollowed, as the spammer may not care to check. Conversely, a spammer can also hit many, many smaller sites to gain a similar amount of PageRank. While this requires more work, smaller sites are more likely to be GhostTowns and therefore unmoderated (and thus inadvertently WildHoneyPots).
Interestingly, robotic attacks are relatively unknown. Most spam is done manually with cheap labour in very undeveloped countries like China (leading to the inevitable RegionalBan of China). This trend should not be counted on. It's merely a matter of time before someone writes a robot that uses Google to search for wikis and then methodically attacks from highest in the rankings to the lowest.
LinkSpam is often ShotgunSpam. ShotgunSpam is best when it is a ForestFire; i.e. spread amongst the widest number of pages. Spam is further improved when they simply change the URLs of a [named link] (e.g. [named link]) in the hopes that PeerReview will overlook the hidden text of the HREF.