In theory a get request sent to a server should not have any side effects and only retrieve some data. In practice implemention is completely up to the developer and their rule is about as useful as putting up an exit sign to prevent people from entering your building.

As others mentioned, GET requests are supposed to be idempotent, i.e., you can send the same request 100 times and get the same response (with no server side-effects) 100 times.

GET requests are also easier to be abused in Cross Site Request Forgery (CSRF) attacks. Modern countermeasures in browsers (like SameSite cookies) will protect cross-origin POST and other state-changing methods, but will largely allow GET requests to go through while carrying session cookies.

I meant read-only there in the sense of mutability, not exfiltration.

Of course, some websites may permit mutations through GET so it’s probably only sensible to use alongside known hosts.