Graphitic carbon nitride compounds were prepared by thermal treatment of C−N−H precursor mixtures (melamine C 3 N 6 H 9 , dicyandiamide C 2 N 4 H 4 ). This led to solids based on polymerized heptazine or triazine ring units linked by −N or −NH− groups. The H content decreased, and the C/ N ratio varied between 0.59 and 0.70 with preparation temperatures between 550 and 650 °C due to increased layer condensation. The UV−vis spectra exhibited a strong π−π* transition near 400 nm with a semiconductor-like band edge extending into the visible range. Samples synthesized at 600−650 °C showed an additional absorption near 500 nm that is assigned to n−π* electronic transitions involving the N lone pairs. These are forbidden for planar symmetric s-triazine or heptazine structures but become allowed as increased condensation causes distortion of the polymeric units. Photocatalysis studies showed there was no correlation between the increased visible absorption due to this feature and H 2 evolution from methanol used for the anodic reaction. In the absence of any cocatalyst the sample synthesized at 550 °C showed 1.5 μmol h −1 H 2 evolution with UV−vis irradiation, but this dropped to ∼0.23 μmol h −1 when the UV spectrum was blocked. Use of a Pt cocatalyst was required to observe H 2 evolution from the other samples. Using a more powerful (300 W) lamp led to higher H 2 production rates (31.5 μmol h −1 ) with visible illumination. We suggest the distorted N sites caused by increased polymerization result in electron/hole traps that counter the photocatalytic efficiency. Issues concerning sample porosity are also present. Photocatalytic O 2 evolution was determined for RuO 2 -coated samples using the 300 W lamp with aqueous AgNO 3 solution as the sacrificial agent. The materials all showed production rates ∼9 μmol h −1 . A highly crystalline compound containing polytriazine structural units ((C 3 N 3 ) 2 (NH) 3 •LiCl) prepared in this study did not show measurable photocatalytic activity.