The distinction is your landing page is only found by people looking for the solution you want to offer. They've already decided they have a problem, and need to fix it.

Contrast that with polling a random sample of your target demographic, where any number of things can spoil the results. People could say they like the idea just to be polite. People could have never thought about the problem you're solving before, but think it's neat. People can misjudge how much they care about it. People can have an enthusiastic initial reaction, but think the idea is dumb on further reflection.

It's the same sort of problem with polling for election results: you can't ask a whole bunch of random people who they're going to vote for and turn the raw results into a pie chart. To get useful data, you have to carefully predict which of your respondents will actually turn up for the election, and which ones aren't going to change their mind between the time you poll them and voting day. And even then, in the end, it's not the pre-election polls that determine the new leader, it's the election itself.

That distinction is what's being presented here; asking a bunch of people is only useful if you're very careful about filtering out the junk answers. You should instead measure people who have already decided, without your prompting, that they have a problem and have decided to seek out a solution for it.