为什么竞价广告的计费价格是竞价排名第2的价格而非竞价排名第1的价格呢？

简单地说，系统创建者怕竞拍者把价格改来改去太麻烦了，就说不使用第一名的价格，而是使用第二名的价格。

为什么使用第一名的价格会带来麻烦呢？

如果在竞价体制中规定，竞价成功的第一名的计费价格为他参与拍卖时的价格。那么会发生下面的场景。如果第一名出价2元且是最高价，第二名的价格是1.80元。那么第一名将会打算在下一轮竞价中，降低一些价格，比如出到1.81元。而第二名可能会打算在下一轮中提高到1.82。第二轮结束后，上一轮的排名顺序被打乱了。这将造成系统的不稳定。

参考资料

www.quora.com

In pay-your-bid auction, if you bid a $2 and the second highest bid is $1.80, you want to reduce your bid down to $1.80 to pay less for the next search. The problem is that the bidder at 1.80 may be willing to pay 1.82 to go up a position, but not $2.00 to do so. That can create cat-and-mouse dynamics, as documented in the paper I mentioned above.

One conjecture I have heard is that the GSP format was born to remove incentives for that kind of behavior and improve efficiency and stability of bids: instead of creating incentives for the highest bidder to go down to 1.81 with his bid, the system does it automatically. 

That does not answer why GSP is still used instead of VCG, but since in a baseline model they are revenue-equivalent, my guess would be that once you find a format that works and is pretty transparent, there are no reasons to redesign the auctions again.