A handheld, spatially offset Raman spectroscopy (SORS) system was successfully used to obtain Surface‐enhanced Raman Scattering (SERS) spectra of fentanyl under simulated field conditions. A series of aqueous fentanyl solutions were prepared with commercially available gold nanoparticle solution, at concentrations ranging from 0.003 to 1697 μM. These SERS spectra were then used to generate two concentration calibration models (via a plot of peak area (1026 cm−1) versus concentration, and quantitative spectral decomposition using partial least squares (PLS1)). For both models, the relationship followed Langmuir adsorption and became non‐linear at concentrations above ~0.2 μM, with a limit of detection (LOD) of approximately 3 nM. The same technique was successfully used to measure fentanyl in the presence of two common “cutting agents,” heroin and glucose, at 1% and 2% fentanyl proportions (w/w). Fentanyl detection was successfully achieved, but mixture interference from the cutting agents prevented a calibration model being generated. Four fentanyl analogues were also investigated—butyrylfentanyl, furanylfentanyl, acetylfentanyl, and ocfentanyl. A concentration calibration model for each species was successfully generated, but differentiation from fentanyl proved more challenging, although several potential diagnostic peaks were identified. These results identified a pathway forward in using handheld equipment for the reliable detection of ultra‐low concentrations of fentanyl and fentanyl analogues via SERS, even when mixed with diluents. However, quantitative detection is negatively impacted in the presence of heroin and glucose. This also provides a starting point for a SERS‐based spectral library of fentanyl analogues, in combination with a range of different diluents.