"This policy is the right one because it's the correct approach." Stated that baldly, nobody is fooled. But circular arguments are rarely that naked — they usually travel through several plausible-sounding steps before quietly arriving back at their own starting assumption, and by then the return is hard to see.
The traditional name is begging the question, from a Latin phrase meaning to assume the very point at issue. Worth noting in passing: in ordinary modern English "begs the question" is now widely used to mean "raises the question," a shift that is too entrenched to fight, so philosophers often prefer "circular reasoning" to avoid confusion.
Detection is a matter of stripping the language. Take the premise and the conclusion, strip both to their plainest wording, and ask whether they say the same thing. Circularity survives by rephrasing: the premise wears different vocabulary than the conclusion, so the identity is disguised. A classic religious-epistemology example makes the shape clear — the text is reliable because it is divinely authored, and we know it is divinely authored because the text says so. Each step is fine on its own; together they support each other and nothing outside the pair.
Two honest complications. First, all reasoning eventually rests on something unproved — you cannot justify every premise without either infinite regress, circularity, or a stopping point, which is the regress problem at the heart of epistemology. The fallacy is not "having assumptions" but presenting a circle as though it were a proof. Second, some circularity is benign: a coherent system of mutually supporting beliefs is not automatically vicious, and coherentists in epistemology defend exactly that structure. The vicious case is when the circle is offered to someone who does not already accept the conclusion — because to them, it gives no reason at all.