google is mankind's greatest engineering feat. i haven't looked into any of it personally, but judging by the replies in this thread, alot of you aren't really aware of basic spam detection much less clustering/classification techniques. if you want to combat the enemy, first you have to understand what it can do. look up naive bayes to get a simple idea of what is possible - P(A|B) = P(A and B) / P(B). machine learning and ai are probably being relied upon heavily for google's spam detection services. if you're going to combat it effectively, your solutions will have to employ similar techniques. a valid point was brought up - machine learning/data mining solutions can be expensive for google to employ. while this is true to a certain extent, it's not entirely true. we're talking about guys that graduated at the top of their class at ivy league schools just to sweep the floors much less work on actual algorithms (exaggeration but you get the point.) -- accept this and you'll do a lot better.