Non-alcoholic fatty liver disease is associated with insulin resistance and dyslipidaemia and can progress to steatohepatitis and cirrhosis. We sought to determine whether dietary fat and saturated fat content alter liver fat in the absence of weight change in an older population. Liver fat was quantified by magnetic resonance spectroscopy before and after 4 weeks on an isoenergetic low-fat/low-saturated fat/low-glycaemic index (LGI) (LSAT: 23 % fat/7 % saturated fat/GI < 55) or a high-fat/high-saturated fat/high-GI (HSAT: 43 % fat/24 % saturated fat/GI>70) diet in older subjects. In the present study, twenty subjects (seven males/thirteen females; age 69·3 (sem 1·6) years, BMI 26·9 (sem 0·8) kg/m2) were randomised to the LSAT diet and fifteen subjects (six males/nine females; age 68·6 (sem 1·8) years, BMI 28·1 (sem 0·9) kg/m2) to the HSAT diet. Weight remained stable. Liver fat decreased significantly on the LSAT diet (median 2·2 (interquartile range (IQR) 3·1) to 1·7 (IQR 1·8) %, P= 0·002) but did not change on the HSAT diet (median 1·2 (IQR 4·1) to 1·6 (IQR 3·9) %). The LSAT diet lowered fasting glucose and total cholesterol, HDL-cholesterol and LDL-cholesterol and raised TAG (P< 0·05), while the HSAT diet had no effect on glucose or HDL-cholesterol but increased total cholesterol and LDL-cholesterol (P< 0·05). Fasting insulin and homeostasis model of insulin resistance did not change significantly on either diet, but the Matsuda index of insulin sensitivity improved on the LSAT diet (P< 0·05). Assignment to the LSAT v. HSAT diet was a predictor of changes in lipid parameters but not liver fat. We conclude that diet composition may be an important factor in the accumulation of liver fat, with a low-fat/low-saturated fat/LGI diet being beneficial.