I will present a detailed study of AGN SED shapes in the optical-near
infrared for 413 X-ray selected Type 1 AGNs from the XMM-COSMOS Survey. We
calculate the mean and dispersion SEDs for this sample (Elvis and Hao et
al. 2012). We divided the sample into quartile bins of redshift, bolometric
luminosity, black hole mass and Eddington ratio and show that the shape of
the mean and dispersion SEDs in each bin do not depend on all these
parameters, which implies an intrinsic mean SED of quasars exists (Hao et
al. 2012a).

The above conclusion depend on the assumption we made in the host galaxy
correction process, which is using the local scaling relationship between
the black hole mass the host galaxy luminosity adding an evolutionary term.
Alternatively, I defined a near-IR/optical index-index ('color-color')
diagram to investigate the mixture of AGN continuum, reddening and host
galaxy contribution (Hao et al. 2012b).

We found that ~90% of the AGNs lie on mixing curves between the Elvis et
al. (1994) mean AGN SED (E94) and a host galaxy, with only the modest
reddening. However, ~10% of the AGNs have weak or non-existent near-IR
bumps, suggesting a lack of the hot dust characteristic of AGNs (Hao et al.
2010, 2011).

The mixing diagram is not only useful in identify outliers but can also
show the proposed 'cosmic cycle' of SMBH and galaxy co-evolution (e.g.
Hopkins et al., 2006) as evolutionary tracks, showing great potential of
revealing how the SMBH and galaxy evolve during cosmic time.