Purpose: The pancreas undergoes erratic and unstable respiratory-induced motion, which hinders respiratory gating and makes dose escalation difficult. In this study, we demonstrate the accuracy and dosimetric impact of respiratory gating using the instantaneous location of fiducial markers (internal gating) compared to conventional techniques for pancreatic tumors during SBRT. Methods: Motion data from 97 pancreatic SBRT treatments were retrospectively analyzed. In each, the daily trajectory of the tumor was reconstructed by calculating a Gaussian probability density function using the location of gold fiducial markers in the CBCT projections. The uncertainty in position for each phase of the respiratory cycle was given as the mean width of the Gaussian PDF. Using the calculated trajectories, we investigated the dosimetric impact of several respiratory motion management strategies, including conventional gating using an external chest block, and internal gating using the fiducial marker location. Results: End-exhale was the most consistent phase, with a mean variance of 0.14±0.14 mm (95% CI 0.04–0.65 mm), whereas max inhale was the least consistent (mean variance 0.34±0.36 mm, 95% CI 0.06–1.46 mm). In a simulated clinical scenario, fiducial-based internal gating using a 2mm superior-inferior window greatly outperformed the conventional approach using external markers (p<0.001), with a mean target D95 of 99.0±2.0%, 95% CI 92.7–100% (conventional gating – D95 96.6±7.4%, 95% CI 68.4–100%). Internal gating also increased the duty cycle from 30% to 45%. Conclusions: Pancreatic motion correlates weakly with chest wall motion, and makes conventional gating less accurate. Fiducial markers are easily discernable in kV imaging, and can provide accurate information regarding the instantaneous tumor location. By acquiring sequential kV images during pancreatic SBRT, the tumor location can be measured, and the treatment beam can be gated using this information. In a clinical dataset of 97 SBRT treatments, this protocol resulted both greater accuracy and a higher duty cycle. This work was funded in part by the National Institutes of Health, the American Cancer Society, and Varian Medical Systems.