Most of the time, they provide no useful model. When they do, they are demonstrably wrong.
For example, there was a guy who had his "theory of gravity" which failed to produce elliptical orbits (IIRC, non-circular orbits were highly unstable).
Other guy tried to explain fusion by drawing electron orbits (not orbitals) of around a nucleus.
IMO, skepticism is the way to go with new theories.
- No useful model? GTFO.
- Model does not fit a lot of experiments? GTFO.
- Model is more narrow and complex than existing theories? GTFO.
- Model does not have measurable effects with current technology? Well, you have to wait or build it yourself.
- Model fits experiments, had successful predictions before experiments, and explains more/unifies other theories? Good, go grab a cookie^W Nobel Prize.
How do you tell the difference between a crackpot and an out of the box thinker? Sometimes it's obvious (I've squared the circle!), but often times it is ambiguous (probably a crackpot, probably not a genius). In this case, I would argue that it is the latter.
99% of all crackpot theories are just that, 1% of the time they are instead paradigm shifts undergoing community backlash. Skepticism is a winning strategy, which is why it is so common, but you can never say for sure unless your crap filters are really precise.