I’ve had the absolute worst time with Akismet lately. So much in fact that I am researching a few ways to replace it. Akismet does do a pretty good job of weeding out the spam and it does learn fairly quickly when you mark something as spam. The problem is in the false positives. If you ask me, any false positives are too many. I know that that’s a perfect scenerio and not really possible, but if a false positive pops up and I de-mark it as spam, then the next time someone comments, they shouldn’t get marked as spam.
Let me show you my latest example. (and the one that has prompted my looking for a new spam solution.)Â P.S. You’ll have to click on the thumbnail to see the full image.
As you can see, there are two comments from TDavid. The first is the one that I had to manually approve today. The second from him is one that I approved earlier this week. There are at least two others further down the list from a month or so ago.
As you can see, it hit the spam comments. WTF? I wouldn’t expect a new comment to hit the spam folder, but from a person who already had one approved? That’s about as buggy as you can get.
Anyone have a good explanation for why they keep ending up in the spam list?
Technorati Tags: spam, akismet, spam folder, approval
Because the spam bot can use the email address of any user that has already posted on your blog? It would all go through..
But as a suggestion, they should allow the comments with the same IP or IP range and email address to go through.
I don’t use akismet. I guess I don’t have the readership to need to do anything other than delete spam myself.
I am not sure but I think Akismet pays more attention to overall spam rating that your personal spam ratings. So if the IP address that person posted from, or that person was marked spam by a number of other blogs it gets marked that way by Aksimet. I agree that is lame – it should essentially whitelist people you have approved. Then it can use some formula if you approve one and then mark 5 more as spam stop whitelisting them or some such…