How verification gets wasted
It is entirely possible to import, verify, read a good signature and have established nothing at all. These are the ways that happens.
Taking the key from the message
If the key and the announcement came from the same source, verifying one against the other proves the source is self consistent. An attacker supplies both and passes this check every time. The key must come from somewhere the message did not.
Comparing a fingerprint you were just given
Same failure in a different position. A fingerprint printed next to the key it authenticates is decoration. The comparison has to be against something you recorded earlier or gathered independently.
Stopping at the trust warning
Reading the not certified line as a failure and abandoning a run that had already succeeded. This is the most common way the step is lost, and it feels like caution while producing the same outcome as never checking.
Verifying then typing the address by hand
A verified announcement establishes which fifty six characters are correct. Retyping them reintroduces exactly the error the verification eliminated. Copy from the verified text.
Verifying once and then trusting the source forever
The check applies to one message. A source that published a correctly signed announcement last month provides no assurance about an unsigned one today, and treating past correctness as ongoing authority is how habits decay.
Doing it in a browser
Handing the check to a website replaces the mathematics with trust in that website. The output looks identical and means nothing. If the tooling is inconvenient, that inconvenience is most of what you are paying for.
The pattern underneath all six
Each one substitutes something convenient for something independent. A key from the same place as the message, a fingerprint printed beside the key, a browser instead of a local tool. Verification only produces information when at least one element came from somewhere the other side did not control, and every mistake here removes that one element.