Barry Ritholtz wrote a blog post knocking the AAII survey as a bad timing tool. He brings up two examples where the bull and bear spread were high, November 27, 2008 and July 20, 2006. While in both cases the market eventually went lower they were actually great times to buy in the short run.
The market bottomed right around November 27,2008 before shooting up 20% into early January 2009. Sure the market went lower into March but that was a very tradable rally. In the case of July 20, 2006 a powerful rally was launched that lasted through January of 2007. Yes, we are lower than that point today but a six month rally is nothing to sneeze at and it took two years before we cracked the July 2006 levels.
The AAII survey is not a long term timing tool and nobody ever claimed it was. Like any other indicator it is not perfect but can provide useful clues.