Abstract In this study, 18 animals were fed two forage-based diets: red clover (RC) and grass silage (GS), in a crossover-design experiment in which methane (CH 4 ) emissions were recorded in respiration chambers. Rumen samples obtained through naso-gastric sampling tubes were analysed by NMR. Methane yield (g/kg DM) was significantly lower from animals fed RC (17.8 ± 3.17) compared to GS (21.2 ± 4.61) p = 0.008. In total 42 metabolites were identified, 6 showing significant differences between diets (acetate, propionate, butyrate, valerate, 3-phenylopropionate, and 2-hydroxyvalerate). Partial least squares discriminant analysis (PLS-DA) was used to assess which metabolites were more important to distinguish between diets and partial least squares (PLS) regressions were used to assess which metabolites were more strongly associated with the variation in CH 4 emissions. Acetate, butyrate and propionate along with dimethylamine were important for the distinction between diets according to the PLS-DA results. PLS regression revealed that diet and dry matter intake are key factors to explain CH 4 variation when included in the model. Additionally, PLS was conducted within diet, revealing that the association between metabolites and CH 4 emissions can be conditioned by diet. These results provide new insights into the methylotrophic methanogenic pathway, confirming that metabolite profiles change according to diet composition, with consequences for CH 4 emissions.