AbstractMetaphysics faces a threat from apparently metaphysics‐friendly non‐epistemic forms of semantics, on which sentences express “worldly” propositions—for example, functions from worlds to truth‐values. The threat goes back to Wittgenstein”s Tractatus Logico‐Philosophicus and is pressed in different forms by various contemporary philosophers. It is that metaphysical claims turn out either trivially true or trivially false, because they express the same proposition as a tautology or contradiction. The problem is shown to generalize to accounts on which sentences express Russellian structured propositions. It applies to logic and mathematics as well as metaphysics. Attempts to solve it by reinterpreting apparently non‐contingent claims as contingent metalinguistic claims or by invoking Fregean semantics are shown to fail. The underlying problem concerns necessary equivalence, not necessary truth, and arises in all domains. To solve it, we must recognize that the form of our representations plays an ineliminable cognitive role that cannot be reduced to their content.