Passive acoustic monitoring (PAM) – an approach that uses autonomous acoustic recording units (ARUs) – can provide insights into the behavior of cryptic or endangered species that produce loud calls. However, extracting useful information from PAM data often requires substantial human effort, along with effective estimates of the detection range of the acoustic units, which can be challenging to obtain. We studied the duetting behavior of pair-living red titi monkeys (Plecturocebus discolor) using PAM coupled with an open-source automated detection tool. Using data on spontaneous duetting by one titi pair, combined with recordings from two Song Meter SM2 ARUs placed within their home range, we estimated that the average source level of titi duets was ~105 dB re 20 μPa at 1 m with an attenuation rate of 8 dB per doubling of distance, and we determined that the detection radius for manual annotation of duets in audio recordings was at least 125 to 200 m, depending on the approach used. We also used a supervised template-based detection algorithm (binary point matching) to evaluate the efficacy of automated detection for titi duets in audio recordings using linear arrays of ARUs within a ~2 km2 area. We used seven titi duet templates and a set of “off-target” howler monkey (Alouatta seniculus) templates to reduce false positive results. For duets with a signal-to-noise (SNR) ratio > 10 dB (corresponding to a detection radius of ~125 m) our detection approach had a recall (the number of all duets that are correctly detected) of 1.0. Performance decreased when including duets with a lower SNR (recall = 0.71, precision = 0.75). The fact that multiple lines of evidence suggest an effective detection radius of 125 to 200 m for titi duets across upland terra firme and seasonally flooded forest lends support to our findings. We suggest that PAM studies of other cryptic but vocally active species would benefit from following similar experimental and analytic procedures to determine an ARU’s effective detection radius and to improve the performance of automated detection algorithms.