By Richard Koman for SiliconValleyWatcher.com

In the aftermath of the AOL leak, people are getting more interested than ever in exactly what search engines are doing with all the information they're collecting. The Times takes a look with a piece today by Saul Hansell, headlined "Advertisers trace paths users leave on the Internet."

Yahoo, it seems, has some complex data modeling in which users are categorized into some 300 groups, based on search history, and then served appropriate ads. Every search company does something similar. As the AOL leak showed, search histories create a portrait of an individual, and targetted searching is an early warning sign of imminent spending, and if you're about to spend money, there are a lot of companies that would like to know about that.

No suprises there. You search, we look at what you search and serve you ads. It gets a little spookier when you realize that cookies allow advertisers to track users across the Web.

“You are no longer targeting people you think will be interested in your product,” said Les Kruger, a senior marketing manager at Cingular. “We know based on your behavior that you are in the market, and we can target you as you bounce around the Internet.”

Tacoda Systems runs its behavior targeting ad software on a network of some 3,500 sites, including the Times. So when you've visited one site, Tacoda can serve you ads related to that subject anywhere you travel on the network. For instance, you visit Cars.com and see car ads when you're on Weather.com.

For its part Google is not impressed with history-based advertising. “Does a user want to see an ad on cars when they are planning their weekend vacation, or do they want to see an ad related to what they are looking at?” asks Google's Tim Anderson.

But there are some impressive anecdotes.

LendingTree, an online loan broker, has experimented with aiming ads at some of the 300 categories of users in Yahoo’s new system, like newlyweds and people who have just moved. It had by far the best results advertising to people who had searched for and read information about borrowing money.

“People aren’t in the market for loans all the time,” said Darren Beck, LendingTree’s vice president for online advertising. “These behavioral targeting models seem like the holy grail. We can find people exactly when they want a loan.”