The most common agents for SEOs are the information ones. These are important because of their power search capabilities. But each agent is given a very small, well-defined purpose. And there are several running in parallel to each other, to achieve an overall task. A.I. uses different methods to build up search agents.
Genetic Algorithms (GA): These are used in random search procedures. They are really efficient on problems that are very difficult to solve. Using them on dynamic data sets is difficult, because the initial pool of data may no longer be valid. They are slow and expensive, but are widely used: stock market prediction, electronic circuit design, traveling salesman problem, international equity strategies, etc.
Artificial Neural Networks (ANN): Used for pattern recognition, and classification tasks in particular. They deal well with imprecise data, as they can classify as "not too sure". There's plenty of different kinds of neural nets: Recurrent, Hopfield Network, Bolzman Mmachine, SVM, Self-Organizing-Maps among others.
Semantic Networks (SN): These show the relationships and concepts. First a graph is created containing nodes which represent concepts and links used to represent reltionships between the concepts. There are six types of Semantic Networks: Categories, Definitional, Assertional, Implicational, Executable, Learning, Hybrid.