Reliable and precise methods capable of unambiguously identifying target analytes in real-world samples are indispensable in various fields, ranging from biological studies and diagnosis to quality control. Among various analytic techniques, nuclear magnetic resonance (NMR) is uniquely powerful as it provides multidimensional data useful for structural analysis at the atomic level. The rich information obtained from various NMR experiments allows one to access not only molecular structures and interactions but also the dynamics and diffusional properties. However, the interpretation of NMR data in the analysis of real-world mixtures can be challenging and is often complicated by the overlap of the NMR resonances of each component. Moreover, the inherently low sensitivity of the NMR technique hampers its implementation in many detections, where the analytes of interest are present at low concentrations. By a combination of heteronuclear NMR, dedicatedly designed sensors, ingenious transduction mechanisms, and powerful NMR pulse sequences, significant advancements were made to conquer these limitations. The present review summarizes the sensing systems that effectively facilitate NMR-based detection with an emphasis on the chemical perspective of sensor design and transduction mechanism. Advances in hyperpolarized sensors to boost the sensitivity of detection will also be included where appropriate.