In order to assess the accuracy with which we can determine the morphologies of AGN host galaxies , we have simulated more than 50,000 ACS images of galaxies with z < 1.25 , using image and noise properties appropriate for the GOODS survey .
We test the effect of central point-source brightness on host galaxy parameter recovery with a set of simulated AGN host galaxies made by adding point sources to the centers of normal galaxies .
We extend this analysis and also quantify the recovery of intrinsic morphological parameters of AGN host galaxies with a set of fully simulated inactive and AGN host galaxies .

We can reliably separate good from poor fit results using a combination of reasonable error cuts , in the regime where \left ( L _ { \mbox { \small { host } } } / L _ { \mbox { \small { PS } } } \right ) > 1 : 4 .
We give quantitative estimates of parameter errors as a function of host-to-point-source ratio .
In general , we separate host and point-source magnitudes reliably at all redshifts ; point sources are well recovered more than 90 % of the time , although spurious detection of central point sources can be as high as 25 % for bulge-dominated sources .
We find a general correlation between Sérsic index and intrinsic bulge-to-total ratio , such that a host galaxy with Sérsic n < 1.5 generally has at least 80 % of its light from a disk component .
Likewise , “ bulge-dominated ” galaxies with n > 4 typically derive at least 70 % of their total host galaxy light from a bulge , but this number can be as low as 55 % .
Single-component Sérsic fits to an AGN host galaxy are statistically very reliable to z < 1.25 ( for ACS survey data like ours ) .
In contrast , two-component fits involving separate bulge and disk components tend to over-estimate the bulge fraction by \sim 10 % , with uncertainty of order 50 % .