Abstract We show that Convolution Neural Networks trained to find strong gravitational lens systems are biased towards systems with larger Einstein radii and large concentrated sources. This selection function is key to fully realising the potential of the large samples of strong gravitational lens systems that will be found in upcoming wide-field surveys. In this paper, we use a CNN and three training datasets to quantify the network selection function and its implication for the many scientific applications of strong gravitational lensing. We use CNNs with similar architecture as is commonly found in the literature. The networks preferentially select systems with larger Einstein radii and larger sources with more concentrated source-light distributions. Increasing the detection significance threshold to 12σ from 8σ results in 50percnt of the selected strong lens systems having Einstein radii θE ≥ 1.04 arcsec from θE ≥ 0.879 arcsec, source radii RS ≥ 0.194 arcsec from RS ≥ 0.178 arcsec and source Sérsic indices $n_{\mathrm{Sc}}^{\mathrm{S}}$ ≥ 2.62 from $n_{\mathrm{Sc}}^{\mathrm{S}}$ ≥ 2.55. The model trained to find lensed quasars shows a stronger preference for higher lens ellipticities than those trained to find lensed galaxies. The selection function is independent of the slope of the power-law of the mass profiles, hence measurements of this quantity will be unaffected. The lens finder selection function reinforces that of the lensing cross-section, and thus we expect our findings to be a general result for all galaxy-galaxy and galaxy-quasar lens finding neural networks.