From mathematician Peter Woit at Not Even Wrong:
Way back in 2005, soon after the emergence of the “String Landscape” and the ensuing debate over whether this made string theory untestable pseudo-science, Cumrun Vafa in response started writing about the “Swampland”. In contrast to the “Landscape” of effective field theories that are low energy limits of string theory, the “Swampland” is the space of effective field theories that are not low energy limits of string theory. One motivation here is to be able to claim that string theory is predictive, since if you can show a theory is in the Swampland, then string theory predicts that doesn’t describe our world.
I wrote a couple blog postings about this back then, see here and here. The situation was rather comical, with Jacques Distler unintentionally making clear one problem with the whole idea. More.
String theory is not something to which evidence makes any difference. It is true because proponents need it to be. They speak for fashionable science, an enterprise they own, along with Darwinism and the multiverse.
See also: Post-modern physics: String theory gets over the need for evidence