Models of star formation predict a maximum surface brightness (flux density per solid angle) that can be achieved (Condon 1992). The figure on the left shows how this depends on the observing frequency (x-axis), and the frequency at which the radio emission has an optical depth of unity (different coloured lines).  Anything above these lines therefore must be due to radio emission from AGN activity.

Reaching the relevant sensitivity to be able to detect high surface brightness requires high resolution - if you think about a source of 1 Jy, its surface brightness is  much lower if spread across a larger area.  By resolving compact components in galaxies, which requires high resolution, we can use their surface brightness to understand if they are generated by AGN activity.

Using the methods above, we can now estimate the AGN contribution to the total radio emission for sources which would otherwise not be identified as AGN which generate radio emission. On the left you can see radio luminosity plotted vs. star formation rate, with some typical cuts for selecting AGN where radio emission is due to AGN activity. The colourscale reveals how we can extend this now to the population where radio emission is currently thought to be dominated from star formation.

For more about this, please see:  Morabito et al., 2025, MNRAS, 536, L32