Sequencing of tumor DNA to detect genetic aberrations is becoming increasingly important, not only to refine cancer diagnoses but also to predict response to targeted treatments. Next-generation sequencing is widely adopted in diagnostics for the analyses of DNA extracted from routinely processed formalin-fixed, paraffin-embedded tissue, fine-needle aspirates, or cytologic smears. PCR-based enrichment strategies are usually required to obtain sufficient read depth for reliable detection of genetic aberrations. However, although the read depth relates to sensitivity and specificity, PCR duplicates generated during target enrichment may result in overestimation of library complexity, which may result in false-negative results. Here, we report the validation of a 23-gene panel covering 41 hotspot regions using single-molecule tagging of DNA molecules by single-molecule molecular inversion probes (smMIPs), allowing assessment of library complexity. The smMIP approach outperforms Sanger and Ampliseq-Personal Genome Machine-based sequencing in our clinical diagnostic setting. Furthermore, single-molecule tags allow consensus sequence read formation, allowing detection to 1% allele frequency and reliable exclusion of variants to 3%. The number of false-positive calls is also markedly reduced (>10-fold), and our panel design allows for distinction between true mutations and deamination artifacts. Not only is this technique superior, smMIP-based library preparation is also scalable, easy to automate, and flexible. We have thus implemented this approach for sequence analysis of clinical samples in our routine diagnostic workflow.