The former is a gamble, but it may be an informed gamble where you already have some idea of how to make it happen. The latter is a lie.
Startups absolutely should be operating in the realm of "It has not been done, but we think we can pull it off. In fact, we are pretty confident of that." They should not be operating in the realm of outright con artistry and fraud.
If you have the right experience, skills, education, personal assets of various sorts and solid mental models, predicting your expected success in a specific domain does not guarantee you will succeed. But it isn't necessarily just BS either.
The challenge is for other people to know which category you fall in. But you should know whether you are genuinely confident or straight up conning people -- barring serious delusion of some sort.