Akismet and False Positives: Using keywords

Akismet (and the rest of them) use algorithms and various procedures to flag things as spam. Their track record is very good, but the problem comes when a poster makes a comment with 3 or more URLs. A lot of times akismet puts it in the moderation queue, but sometimes it goes directly to the akismet spam bin (which people usually don’t inspect closely because of the volume).

It’s painful (even impossible) to go through individual akismet comments, but here’s a time saving method if you ever want to.

Use the search box feature to type in certain popular keywords for your blog: For example, on teleread you could use these three or four terms:

1. ebook
2. publishing
3. novel
4. publishing
5. author
6. copyright

and this would probably be enough to spot false positives. It’s somewhat likely that a legitimate comment would include one of these words, while it’s very unlikely that a spam comment would include them. A search using these terms would take 30 seconds, and it would allow you to inspect akismet comments more efficiently.

The problem with this shortcut would come if comment spam were customized towards individual blogs. For example, perhaps a commenting engine would analyze teleread and conclude that certain types of URL’s are more appropriate (or it would recycle legitimate comments from the blog with the spam to evade detection). That would seem to require too much intelligence (for the time being).