Nobody consciously times a website. But everyone feels it. In the first two seconds a page is loading, a visitor isn't thinking "this is slow", they're thinking something closer to "is this thing broken". By five seconds, most of them have already left, and the ones who stayed have already formed an opinion about the business behind it, before a single word of the actual pitch has loaded.
That opinion isn't really about speed. It's about competence. A slow site reads the same way a business card printed on flimsy stock does: not fatal on its own, but a small, specific signal that something wasn't finished with care. Visitors don't reason their way to that conclusion. They just feel it, and they carry the feeling into everything that follows.
The frustrating part is how rarely load time is actually a hard technical problem. Most slow sites aren't slow because the work being done is heavy. They're slow because of accumulation: a page builder rendering more than it needs to, images shipped at their original size instead of the one the browser actually displays, a script loaded for one feature buried three pages deep, fonts pulled from four different places. None of that is unavoidable. All of it is fixable, usually without touching the design at all.
This is also why "add a loading spinner" is the wrong fix. A spinner doesn't make the wait shorter. It just tells the visitor you know they're waiting, which is a strange thing to be proud of. The real fix is almost always structural: fewer requests, smaller assets, less shipped than rendered. It's less visible work than a redesign, and it matters more.
Speed is one of the few things about a website a visitor experiences before they experience anything else you've made. It's worth treating like the first impression it actually is.