Spam detection poison is content that makes spam detection databases get corrupt and leads to “false Positives”. If you feed certain types of email into a Bayesian database it can negatively impact the ability to detect real spam. The spammers are well aware of this!
If you discover that more and more email is landing in your spam bucket it might be wise to pay more attention to WHAT you are indicating is SPAM.
An email with little to no content or content that would normally not be considered spam, if it was sent by someone you know, is exactly the type of spam that poisons spam databases.
These emails generally contain familiar greeting and words that would be completely in line with and old friend getting in touch or a legitimate business inquiry.
For emails like this we take the time to track down the sender and review the amount of spam on our mail servers from the IP address. Then we block the IP address instead of “Learnin Spam” on the content.
If spam is originating from multiple IP addresses in the same IP range we block the entire range.