Plant‐ and alga‐associated bacterial communities are generally described via 16S rDNA metabarcoding using universal primers. As plastid genomes encode 16S rDNA related to cyanobacteria, these data sets frequently contain >90% plastidial sequences, and the bacterial diversity may be under‐sampled. To overcome this limitation we evaluated in silico the taxonomic coverage for four primer combinations targeting the 16S rDNA V3‐V4 region. They included a forward primer universal to Bacteria (S‐D‐Bact‐0341‐b‐S‐17) and four reverse primers designed to avoid plastid DNA amplification. The best primer combination (NOCHL) was compared to the universal primer set in the wet lab using a synthetic community and samples from three macroalgal species. The proportion of plastid sequences was reduced by 99%–100% with the NOCHL primers compared to the universal primers, irrespective of algal hosts, sample collection and extraction protocols. Additionally, the NOCHL primers yielded a higher richness while maintaining the community structure. As Planctomycetes, Verrucomicrobia and Cyanobacteria were underrepresented (70%–90%) compared to universal primers, combining the NOCHL set with taxon‐specific primers may be useful for a complete description of the alga‐associated bacterial diversity. The NOCHL primers represent an innovation to study algal holobionts without amplifying host plastid sequences and may further be applied to other photosynthetic hosts.