This paper describes the implementation and performance of CsI(Tl) pulse shape discrimination for the Belle II electromagnetic calorimeter, representing the first application of CsI(Tl) pulse shape discrimination for particle identification at an electron-positron collider. The pulse shape characterization algorithms applied by the Belle II calorimeter are described. Control samples of $gamma$, $mu^+$, $pi^pm$, $K^pm$ and $p/bar{p}$ are used to demonstrate the significant insight into the secondary particle composition of calorimeter clusters that is provided by CsI(Tl) pulse shape discrimination. Comparisons with simulation are presented and provide further validation for newly developed CsI(Tl) scintillation response simulation techniques, which when incorporated with GEANT4 simulations allow the particle dependent scintillation response of CsI(Tl) to be modelled. Comparisons between data and simulation also demonstrate that pulse shape discrimination can be a new tool to identify sources of improvement in the simulation of hadronic interactions in materials. The $K_L^0$ efficiency and photon-as-hadron fake-rate of a multivariate classifier that is trained to use pulse shape discrimination is presented and comparisons are made to a shower-shape based approach. CsI(Tl) pulse shape discrimination is shown to reduce the photon-as-hadron fake-rate by over a factor of 3 at photon energies of 0.2 GeV and over a factor 10 at photon energies of 1 GeV.