Remote cameras have become a widespread data‐collection tool for terrestrial mammals, but classifying images can be labor intensive and limit the usefulness of cameras for broad‐scale population monitoring. Machine learning algorithms for automated image classification can expedite data processing, but image misclassifications may influence inferences. Here, we used camera data for three sympatric species with disparate body sizes and life histories – black‐tailed jackrabbits (Lepus californicus), kit foxes (Vulpes macrotis), and pronghorns (Antilocapra americana) – as a model system to evaluate the influence of competing image classification approaches on estimates of occupancy and inferences about space use. We classified images with: (i) single review (manual), (ii) double review (manual by two observers), (iii) an automated‐manual review (machine learning to cull empty images and single review of remaining images), (iv) a pretrained machine‐learning algorithm that classifies images to species (base model), (v) the base model accepting only classifications with ≥95% confidence, (vi) the base model trained with regional images (trained model), and (vii) the trained model accepting only classifications with ≥95% confidence. We compared species‐specific results from alternative approaches to results from double review, which reduces the potential for misclassifications and was assumed to be the best approximation of truth. Despite high classification success, species‐level misclassification rates for the base and trained models were sufficiently high to produce erroneous occupancy estimates and inferences related to space use across species. Increasing the confidence thresholds for image classification to 95% did not consistently improve performance. Classifying images as empty (or not) offered a reasonable approach to reduce effort (by 97.7%) and facilitated a semi‐automated workflow that produced reliable estimates and inferences. Thus, camera‐based monitoring combined with machine learning algorithms for image classification could facilitate monitoring with limited manual image classification.

Context Camera trapping is an effective tool for cost-efficient monitoring of species over large temporal and spatial scales and it is becoming an increasingly popular method for investigating wildlife communities and trophic interactions. However, camera trapping targeting rare and elusive species can be hampered by low detection rates, which can decrease the accuracy and precision of results from common analytical approaches (e.g., occupancy modeling, capture-recapture). Consequently, researchers often employ attractants to increase detection without accounting for how attractants influence detection of species among trophic levels. Aims We aimed to evaluate the influences of a commonly used non-species-specific olfactory lure (i.e. sardines) and sampling design on detection of four species (i.e. bobcat [Lynx rufus], coyote [Canis latrans], raccoon [Procyon lotor], and eastern cottontail [Sylvilagus floridanus]) that represented a range of foraging guilds in an agricultural landscape. Methods We set 180 camera stations, each for ∼28 days, during the summer of 2019. We set cameras with one of three lure treatments: (1) olfactory lure, (2) no olfactory lure, or (3) olfactory lure only during the latter half of the survey. We evaluated the influence of the lure at three temporal scales of detection (i.e. daily probability of detection, independent sequences per daily detection, and triggers per independent sequence). Key results The lure tended to positively influence detection of coyotes and raccoons but negatively influenced detection of bobcats and eastern cottontails. The influence of the lure varied among temporal scales of detection. Conclusions Scent lures can differentially influence detection of species within or among tropic levels, and the influence of a scent lure may vary among temporal scales. Implications Our results demonstrate the importance of evaluating the influence of an attractant for each focal species when using camera data to conduct multi-species or community analyses, accounting for variation in sampling strategies across cameras, and identifying the appropriate species-specific temporal resolution for assessing variation in detection data. Furthermore, we highlight that care should be taken when using camera data as an index of relative abundance (e.g. as is commonly done with prey species) when there is variation in the use of lures across cameras.