Abstract – The horizontal distribution of fish in lakes varies with species, age, predation risk, vegetation coverage and water clarity. Although sporadic data are available for specific lakes, little is known about how habitat distribution generally changes along a gradient in nutrients and clarity. Here we analyse littoral and pelagic fish catch data for 34 Danish lakes (covering 94 lake years) with contrasting total phosphorus (TP) concentrations and mean depth levels (but mainly eutrophic and shallow). Electrofishing was conducted during day along the shore or along the edge of the reed belt if such existed. Gill nets were set overnight for 18 h in open water. We calculated a littoral proportion as: LITPRO = (100 × number caught by electrofishing in the near‐shore area)/(number caught by electrofishing in the near‐shore area and nets in the open water). At all summer TP concentrations LITPRO was high for several ‘littoral’ species such as gudgeon (Gobio gobio), rudd (Scardinius erythrophtalmus) and pike (Esox lucius) and low for more typical pelagic species like pikeperch (Sander lucioperca), ruffe (Gymnocephalus cernuus) and smelt (Osmerus eperlanus). However, for many of these species including the dominant roach (Rutilus rutilus) and perch (Perca fluviatilis) LITPRO increased with TP or lake water chlorophyll a (CHLA) for fish >10 cm, suggesting a more littoral distribution with increasing turbidity. Exceptions are pikeperch and bream (Abramis brama). For bream, LITPRO increased significantly with mean lake depth only. For fish ≤10 cm LITPRO showed only a weak relationship to TP, CHLA or depth. In accordance with the TP‐dependent distribution, major changes occurred in the fish community structure during the monitoring period following nutrient loading reduction. A comparison of LITPRO from the first part of the monitoring period (1989–1993) with a subsequent period (1997–2003) for four abundant species showed a significant decline in LITPRO over time for large‐ and small‐sized roach and perch and large rudd. This suggests a fast response in fish distribution to improved water quality. It further demonstrates that fish monitoring in lakes should cover both littoral and pelagic habitats to describe the recovery process.