A major obstacle in the culture-independent estimation of the abundance of bacteria associated with plants is contamination with plant organelles, which precludes the use of universal rRNA bacterial primers in quantitative PCR applications. We present here a PCR-based method that allows a priori determination of the degree of chloroplast and mitochondrial contamination in DNA samples from plant environments. It is based on differential digestibility of chloroplast, mitochondrial and bacterial small subunit rRNA gene amplicons with the restriction enzymes AfeI and BbvCI. Using this method, we demonstrated for field-grown lettuce plants that even a gentle washing protocol, designed to recover the microbial community and its metagenome from the leaf surface, resulted in substantial contamination with chloroplast DNA. This finding cautions against the use of universal primer pairs that do not exclude chloroplast DNA from amplification, because they risk overestimation of bacterial population sizes. In contrast, contamination with mitochondrial 18S rRNA was minor in the lettuce phyllosphere. These findings were confirmed by real-time PCR using primer sets specific for small subunit rRNA genes from bacteria, chloroplasts, and mitochondria. Based on these results, we propose two primer pairs (534f/783r and mito1345f/mito1430r) which between them offer an indirect means of faithfully estimating bacterial abundances on plants, by deduction of the mito1345f/mito1430r-based mitochondrial count from that obtained with 534f/783r, which amplifies both bacterial and mitochondrial DNA but excludes chloroplast. In this manner, we estimated the number of total bacteria on most leaves of field-grown lettuce to be between 10 5 and 10 6 g − 1 of leaf, which was 1–3 orders of magnitudes higher than the number of colony-forming units that were retrieved from the same leaf surfaces on agar plates.