High-cycle fatigue (HCF) tests were conducted on samples fabricated by two powder-bed additive manufacturing techniques. Samples were tested with as-produced surfaces and after various non-contact surface improvement treatments. Ti-6Al-4V samples were made using both electron beam melting (EBM) and selective laser melting (SLM), while Inconel 625 was fabricated using SLM. Ti-6Al-4V was treated with a commercial chemically accelerated vibratory polishing process, with target material removal of approximately 200 µm from each surface for EBM samples and 100 µm for SLM samples. This technique led to increases in both the number of cycles before failure at a given loading condition and endurance limit (at 107 cycles) compared to samples with as-produced surfaces. The results are interpreted as the reduction in elastic stress concentration factor associated with surface defects where fatigue cracks initiate. SLM 625 was treated with both an abrasive polishing method and laser surface remelting. Both methods led to improvements in surface roughness, but these did not lead to improvements in fatigue properties of SLM 625. For abrasive polished samples, the combination of improved measured surface roughness without fatigue property enhancement suggests that surface material is removed, but the roots of surface defects, where fatigue cracks initiate, were left intact. For laser treatment, the remelted surface layer retained a rapidly solidified microstructure that did not increase the number of cycles before crack initiation even though the surface was smoother compared to the surface prior to polishing.