In practice, no: a phone camera’s autofocus distance is not reliable enough for precise close-range measurement.

Your geometric approach is basically a similar-triangles/magnification calculation, and the sample numbers are in the right ballpark, but small errors in focal length, sensor size, pixel measurement, focus position, lens distortion, and where “distance” is referenced from can easily create noticeable error. Smartphone camera modules are optimized to make pleasing photos, not act as calibrated measuring instruments.

That also explains why the autofocus-reported distance may differ from a tape-measured subject distance: it is often only an approximate focus estimate, not a precision metrology output.

So the short answer is that this is the wrong tool for accurate measurement. If you need precise distance or size, use a sensor designed for measurement rather than relying on camera AF metadata.

If you still want to estimate size from images, treat it as an approximate vision/calibration problem and calibrate the exact device/camera setup you’re using. But you should expect device-to-device variation, so a calibration may not transfer accurately across all phones of the same model.