I believe that the "output followed the input exactly", at least at first sight.
But input resistance is only typically 28 kohm, prone to type and temperature variation. To keep the AMC1x00 accuracy, the source resistance must be respectively low. If you consider e.g. +/- 10% Rin variations, the source resistance must be in a 1k range to limit the additional gain error to 0.5 %. If you can give the voltage divider respecitively low resisrtance, it's no problem.