Large wind turbines (WTs) can produce noise that is observable at ground or atmosphere receivers up to 10 s of km from the source. Like other machinery with rotating blades (e.g., helicopters and large fans), WTs noise is often characterized by sharp spectral peaks at the blade passing frequency and its integer harmonics. Noise with these spectral characteristics, or tonal noise (TN), can be indistinguishable from typical seismic background noise because it is easily masked in noise studies after traditional smoothing algorithms and can therefore go undetected. Here, we describe and quantify a methodology to detect TN in seismic records by estimating spectral sequences in background noise. To test the performance of this algorithm, we analyze several years of seismic data recorded on two stations (AMTX and NATX) within the United States National Seismic Network (USNSN) in Texas. Station AMTX has a high concentration of WTs within 100 km and shows persistent TN detections with multiple well‐defined fundamental frequencies between 0.7 and 1 Hz. NATX has the closest WT deployment at approximately 250 km and does not display any persistent detection, which provides a test for the algorithm in the absence of signal. Monthly detections at station AMTX follow the multiyear trend of statewide wind‐generated electric power, which supports the assumption of the TN to be related to the operation of WTs. Our method will likely advance future monitoring challenges because WT proliferation spreads globally and such signatures will become increasingly present in seismic data.