The limitations of the alkane technique in estimating the diet components of herbivores call for the introduction of new diet composition markers. Recently, long-chain alcohols (alcohols) and long-chain fatty acids (acids) have received the most attention and show great potential, when combined with alkanes, to estimate composition of complex diets. In the current study, faecal recoveries of alcohols and acids were determined in sheep in four different live weight groups fed three herbage species, either Leymus chinensis, L. dasystachys or Elymus sibiricum. Analysis of variance (ANOVA) was used to examine the effects of herbage species and live weight of sheep on faecal recoveries of individual alcohols and acids. Further, an indoor experiment with six sheep fed a diet of equal proportions, on dry matter (DM) basis, of three herbages was performed, allowing to assess the accuracy of alcohols and/or acids in combination with alkanes, to estimate diet composition. A one-sample t-test was carried out to test the accuracy of these estimates. Results of the first experiment indicated that the faecal recoveries of alcohols and acids were significantly affected by herbage species (P < 0.05). While the effects were significant or near significant for the faecal recoveries of some alcohols (C24-ol, C30-ol and C26-ol) (P ⩽ 0.05), no effect of live weight on faecal recoveries of acids was observed (P > 0.05). Therefore, adjustments based on diet-specific faecal recoveries might improve diet composition estimates. This was illustrated by the results of the second experiment. The diet composition estimated from alcohols or all combinations of alcohols with other marker types, after diet-specific correction of faecal recoveries, did not significantly differ from the actual composition (P > 0.05). However, using acids as additional markers resulted in poorer diet composition estimates. This study confirmed the utility of alcohols, combined with alkanes, as markers to estimate composition of complex diets. Although corrections based on mean faecal recoveries, average over animals and diets, resulted in some accuracy loss, results were still satisfactory and better than without recovery correction.