Utilization of long-chain alcohols (LCOH) as diet-composition markers in sheep consuming six diets composed of improved pasture species (Lolium perenne and Trifolium repens) with heather–gorse components (Erica spp., Calluna vulgaris and Ulex gallii) was evaluated. Twenty-four adult cross-bred sheep were housed in individual stalls. Diet composition was estimated from LCOH concentrations, combined or not with alkanes and long-chain fatty acid (LCFA) data using least-squares procedures. Prior to calculations, faecal concentrations were corrected using mean treatment (faecal recovery 1, FR1) recoveries and mean recoveries across diets (FR2). Estimates were compared with those obtained without faecal correction (FR0) and known values. Large differences between plant species and plant parts were found in LCOH patterns and total LCOH concentrations. LCOH provided complementary information to that given by alkanes and LCFA. Even-chain LCOH comprised the largest fraction, representing on average 0·894 of total concentrations. Faecal recovery was incomplete and tended to increase with carbon-chain length (CCL) in a curvilinear manner (FR = −7·872 + 0·580 × CCL − 0·010 × CCL2; P < 0·001; r2 = 0·752). Diet composition influenced (P < 0·001) LCOH recovery, although variability within (CV of 9·0%) and between (CV of 13·3%) diets was low. Accuracy of estimates was influenced (P < 0·001) by faecal-correction method and markers used, and the best estimate was obtained combining LCOH and alkanes. Improvement in estimate accuracy can be achieved if suitable correction of LCOH faecal concentrations is performed prior to calculations. Results indicated that LCOH showed lower dependence than alkanes and LCFA on the use of accurate faecal-correction data.