Knowing which arthropod and plant resources are used by generalist predators in agroecosystems is important to understand trophic interactions and the precise ecological role of each predatory species. To achieve this objective, molecular approaches, such as the use of high-throughput sequencing (HTS) platforms are key. This study develops a multi-primer metabarcoding approach and explores its suitability for the screening of the most common trophic interactions of two predatory species of arthropod with contrasted morphology, Rhagonycha fulva (Coleoptera: Cantharidae) and Anthocoris nemoralis (Hemiptera: Anthocoridae) collected in an organic peach crop. To save time and cost in this metabarcoding approach, we first evaluated the effect of two different predator-pool sizes (10 and 23 individuals of the same species), as well as the performance of using one or two primer pairs in the same library. Our results show that the analysis of 23 individuals together with the use of two primer pairs in the same library optimizes the HTS analysis. With these best-performing conditions, we analyzed whole bodies of field-collected predators as well as the washing solutions used to clean the insect bodies. Results showed that we were able to identify both, gut content (i.e. diet) as well as external pollen load (i.e. on the insects body), respectively. This study also demonstrates the need of washing predatory insects prior to HTS analysis when the target species have a considerable size and hairy structures. This metabarcoding approach has a high potential for the study of trophic links in agriculture, revealing both expected and unexpected trophic relationships.