Detection of Onchocerca volvulus larvae in vector populations is of prime importance in the assessment of the effectiveness of onchocerciasis control programmes. Traditionally, detection of larvae is attained by the dissection of flies, but this time-consuming method cannot easily discriminate between species of Onchocerca. The genome of all Onchocerca species has a unique 150 bp repeat, which can be amplified by PCR, and O. volvulus-specific DNA probes can detect these products by Southern blot (SB). This study optimizes a PCR/SB assay, and compares it with fly dissection to estimate the prevalence (p) and intensity of infection (m) in the local vector population of a Mexican community that has become hypoendemic as a result of 7 years of treatment with ivermectin and nodulectomy. The PCR detected 1 infected fly in a pool of 99 uninfected flies, but the optimal pool size was 50 flies. At the community level, 1 out of 10550 flies was positive (p=0·0095%, 95% confidence intervals CI=0·00024–07·05280%; m=0·00027 larvae/parous fly, CI=−0·00026–0·00081) by PCR, and 4 out of 10772 flies (p=0·0371%, CI=0·01012–0.09505%; m=0·00107 larvae/parous fly, 95% CI=0·00002–0·00212) by dissection (observed m=0·0005). Both methods produce statistically similar estimates of the prevalence and intensity, indicating that pool screening is a viable alternative for entomological surveillance in areas where the intensity of transmission is becoming extremely low as a result of control interventions.