How technology detection actually works
Websites leak what they are built with, mostly through laziness rather than intent. A CMS writes a generator tag because that is the default. A server announces itself in the Server header. A framework leaves its hydration payload in the page. A theme loads its scripts from paths that name the platform. Detection is nothing more exotic than reading these and knowing which belong to what.
The consequence worth understanding: detection quality depends entirely on what the site chooses to send. A default WordPress install is identifiable four different ways; a hardened one behind a strict proxy may be identifiable zero ways. The confidence ladder exists because these are different situations and a tool that reports both as flat facts is flattening exactly the distinction that matters.
That is also why absence proves nothing. A report with an empty language row does not mean the site has no backend language. It means the backend declined to introduce itself, which well-run sites deliberately do.