Unraveling how reactive facets promote photocatalysis at the molecular level remains a grand challenge, while identification of the reactive facets can provide guidelines for designing highly efficient photocatalysts and unravelling the microscopic mechanisms behind them. Recently, a series of polytriazine imides (PTIs) was reported with highly crystalline structures; all had a relatively low photocatalytic activity for overall water splitting. Here, high-angle annular dark-field scanning transmission electron microscopy, energy dispersive spectroscopy mapping, and aberration-corrected integrated differential phase contrast imaging were used to study PTI/Li+Cl− single crystals before and after in situ photodeposition of co-catalysts, showing that the prismatic {10
$$\bar 1$$
0} planes are more photocatalytically reactive than the basal {0001} planes. Theoretical calculations confirmed that the electrons are energetically favourable to transfer toward the {10
$$\bar 1$$
0} planes. Upon this discovery, PTI/Li+Cl− crystals with different aspect ratios were prepared, and the overall water splitting performance followed a linear correlation with the relative surface areas of the {10
$$\bar 1$$
0} and {0001} planes. Our controlling of the reactive facets directly instructs the development of highly efficient polymer photocatalysts for overall water splitting. Unlike with inorganic photocatalysts, the facet-dependent reactivity of conjugated polymers remains elusive. Now, the authors provide molecular-level insights on the reactive facets of crystalline poly(triazine imide) intercalated with LiCl and achieve a remarkable improvement in its overall photocatalytic water splitting activity.