This paper aims to shed some light on the role of the direct, or nonsequential, ionization channel in the regime in which the sequential channel is open in two-photon double ionization (TPDI) of helium. In this regime the sequential channel dominates any direct contribution unless the laser pulse is of very short duration, in which case their distinction is hard to draw. Based on both a simple model and full solutions of the time-dependent Schrödinger equation, we aim to provide evidence of direct double ionization by identifying a term proportional to the pulse duration in the double ionization yield. Indeed, such a term is identified in the energy-differential yield. When it comes to the total double ionization probability, however, it turns out that the net first-order contribution is negative. The nature of the negative first-order contribution is discussed, and we argue that it is of correlated origin.