Self-organization is unplanned organization that emerges from an open system of interacting components. The system can be thought of as lifting itself up by its boot straps.
This is the way a self-organizing Web site can be developed.
A Web site is the visible part of a wider system that includes: visitors to a Web site; the people who make changes to the Web site; the people who choose the information that appears on the Web site; search engines; the whole of the world wide Web and the Internet.
All of these various factors have an influence on the way in which a a Web site will self-organize.
It is impossible to even begin to try to understand how all these different influences have an effect on the organization that emerges.
This is why it is necessary to use an evolutionary strategy and develop the Web site over a series of rebuilds.
With an evolutionary strategy, you don't plan or predict how the system will perform. You observe what the system is doing then make changes to see if it is performs better. Using a controlled trial and error process, the system has to be steered along a path of increasing order and organization.
This is how Nature achieves self-organization and although it might seem to be a crazy way to design a Web site it is surprizingly effective.
It may be difficult to see how a Web site can be self-organizing, if it is dependent upon humans making decisions and making changes. But, these people are themselves components of the system. They are not planning what the system will do, but responding to what it is doing. Their decisions and changes are influenced as much by the system as their actions are influencing the system.
For a fuller appreciation of such mutual interdependence, it is worth reading Ethan Decker's excellent tutorial Self-Organizing Systems: A Tutorial in Complexity