We discuss the limits of ground-based astrometry with adaptive optics based on experiments using the core of
the Galactic globular cluster M5. We have recently achieved ⪅ 100microarcsecond astrometric precision and
accuracy at the Hale 200-inch telescope. Here we apply the same experimental design considerations and optimal
estimation technique to explore the astrometric precision of the Keck II telescope. We find that high-precision
astrometry at ≈ 50 microarcsecond level is possible at Keck in 20 seconds. We discuss the potential of differential
astrometry for current and next generation large aperture telescopes based on these results.