1. Aquatic herbivores typically have much higher concentrations of nutrients (e.g. N and P) in their tissues than there is in the food they eat. These stoichiometric differences can cause herbivores to be limited by the elemental quality of their food, which could affect, in turn, the structure of consumer communities and even alter key ecosystem processes. 2. In streams and in the littoral zone of shallow lakes, periphyton is an important food resource for benthic animals. Studying the elemental composition of periphyton may help us to understand food-web structure, and any reciprocal effect of this structure on periphyton stoichiometry. 3. To understand how alterations in the food-web structure affect the elemental composition of periphyton in a eutrophic lake, we carried out a long-term experiment (14 months) in large-scale mesocosms (40 m3), in which we manipulated food-web structure, and which were dominated either by planktivorous fish (Rutilus rutilus) or herbivorous invertebrates (without fish). Periphyton was sampled monthly at three depths (0.5, 1.5 and 2.5 m) to determine its biomass and elemental composition (C ⁄N⁄P ratio). Food-web structure, physical and chemical parameters were monitored throughout the experiment. 4. Fish had indirect positive effect on periphyton biomass, leading to twofold higher levels than in herbivore-dominated mesocosms. This result was probably due to control of benthic consumers by fish, suggesting a strong top–down control on periphyton by their consumers in fishless enclosures. 5. The elemental ratios C⁄P and C⁄N were lower in deep water in both treatments, mainly mediated by light availability, in accordance with the light ⁄ nutrient ratio hypothesis. These ratios were also lower in fishless treatments, probably due to increases in inorganic nutrient availability and grazing pressure in herbivore-dominated systems. During winter, periphyton elemental composition was similar in both treatments, and was unrelated to inorganic nutrient availability. 6. These results indicate that any alteration of food-web structure in lakes, such as in biomanipulation experiments, is likely to modify both the biomass and elemental quality of periphyton. Resultant effects on the consumers of periphyton and macrophytes could play a key role in the success of biomanipulations and should be taken into account in further studies.