Purpose: To identify predictors of best corrected visual acuity (BCVA) in eyes with foveal-sparing geographic atrophy (GA) secondary to age-related macular degeneration (AMD). Methods: Best corrected visual acuity (Early Treatment Diabetic Retinopathy Study charts); serial fundus autofluorescence; and near-infrared reflectance images of patients participating in the FAM (NCT00393692) and DSGA (NCT02051998) studies were analyzed. The sizes of GA and spared fovea, and the minimal linear dimension of intact retinal pigment epithelium (``bridge'') between the residual foveal island and the surrounding retina were quantified and associations with BCVA were assessed by local regression curves and mixed effects models. Results: A total of 65 eyes (51 patients, aged 75.68 $±$ 8.41 years) were included. Median time between baseline and last visit with detectable foveal sparing was 18 (quartiles: 12, 33) months. Median BCVA was 0.30 (0.20, 0.52) logMAR at baseline and 0.4 (0.3, 0.7) logMAR at follow-up. Local regression curves suggested no linear association of BCVA with GA size, sparing size or bridge size. Most contrasting values for BCVA were observed for >1.5 mm2 foveal-sparing size and for 400 $μ$m bridge size. Employing these values as cutoff levels, mixed effects modeling revealed that both anatomic parameters, but not time, significantly impacted BCVA. Conclusions: During the review period eyes with foveal-sparing GA were likely to maintain the baseline BCVA. There was no linear correlation of BCVA with foveal-sparing size. Yet, BCVA was worse if the spared foveal area was <1.5 mm2 or if the bridge was smaller than 400 $μ$m in width. These findings add to the understanding of the natural history of foveal-sparing GA and may support future clinical trial designs.