In this talk I introduce the notions of exactly modalized verification (and falsification). In Kit Fine’s approach to exact truthmaker semantics, sentences have what I call categorical exact verifiers, determined purely by the exact verifiers of their subsentences. This however does not take into account the relations of incompatibility and exhaustiveness that might hold between the states that verify those subsentences. Consider p-or-not-p: this sentence is categorically exactly verified by exact p-verifiers and by exact p-falsifiers, but the fact that this particular sentence is always true is due to the relation between those verifiers rather than to the verifiers themselves (the fact that it is impossible that a state is both incompatible with all p-verifiers and with all p-falsifiers). Fine proposes loose verification as a modal notion that does take into account those relations. Unfortunately Fine’s loose verification is fundamentally non-exact: it is not more fine-grained than ordinary notions of verification found in traditional truth-functional semantics and so it misses the benefits of turning toward truthmaker semantics in the first place. We propose an alternative notion of modalized verification, called exactly modalized verification, that does guarantee exactness, in the sense that all exactly modalized verifiers are wholly relevant for the sentences they verify. We use relevance logic to formally capture the notion `relevant’ in this characterization of exactness. The resulting semantics is such that the empty state is an exactly modalized verifier of each tautology and of each exactly valid sequent (a sequent is exactly valid if it is valid and all of its formulas contribute to the validity). We argue that this feature is philosophically interesting: it enables characterizing tautologies as analytic propositions (i.e. propositions that do not need substantial worldly states for their truth), while still maintaining that they have non-trivial content. We will also discuss an alternative proposal for modalized verification, called sub-loose verification, proposed by Fine in a recent reaction to our work. We argue that, while more fine-grained than loose verification, this notion is not exact in the above sense and obfuscates important distinctions that can be made by proper exact verification, such as the distinction between (p-and-q)-or-(r-or-not-r) and (p-and-q)-or-(p-or-q)-or-(r-or-not-r).