Seed uses a Markov Chain text generator to create dummy text to populate entry fields.
The text generated is random, based on seed text included within the addon. When using the text-generation you may see some english looking words, even sentences, however all the text being semi-randomly generated on the fly.
Note: This next bit gets mathematical, and is not required to use Seed in the slightest. It's only included here because it's interesting (to me).
A Markov Chain is a basically a state-less random process where the next step in the process is dependent on the last step. They're perfectly suited to random text generation.
In seed we supply our generator a set of predefined english text (in our case we're using a collection of writings by Kant). The generator takes this text and generates a statistical table of the characters within. From that we can get a weighted probability of which characters are generally found together.
From those weighted probabilities we can generate pieces of text of any length and have the output look like semi-readable english.
The bonus of using a Markov Chain based generator for this purpose is that we can easily switch up the perceived language of the generated content by simply supplying a different set of seed text. This could then be used to generate multi-language dummy text. (This isn't an option in Seed v0.9, but may be presented in future versions).
More information on the wonderful world of Markov Chains is available on Wikipedia (also more general article on Finite State Machines is relevant as they're the general case). Mathworld has the formal definition, or this explanation on StackOverflow is quite useful for understanding the concept.