Here's an interesting piece on how web searches for a company increase just before trading in that stock spikes. That should be a no-brainer; if a stock is being researched, some folks will look to buy or sell based on what they're finding out and the research will likely be a leading indicator of activity.
However, that seems to skip past the folks at MIT-
That's surprising because Cristelli and co say that most webusers check each stock of interest just once per month. This implies that they are not expert traders. So the effect emerges from the collective but uncoordinated activity of many inexpert users--a kind of wisdom crowds.
Checking one a month might be a fair pattern for a stock that is either in your portfolio or is being tracked for possible purchase, but expert traders might well do research into a stock before trading; if something came up on their radar via any variety of screening tools, they might go on the Web to see what is making the company tick.
Note that this just tracks trading activity. It doesn't say that the stock will go up or down, just that extra shares will be traded.