I gave it a photo I had Stable Diffusion 1.4 generate from the prompt "avatar for saurik". If you dig through the CLIP database, you will find that the model was trained on a ridiculously large number of copies of my Twitter profile photo due to it being included when people screenshot popular tweets I've posted (which, notably, also means that it is rather low resolution).

https://www.stableattribution.com/?image=e89f1e94-067b-4ab8-...

https://pbs.twimg.com/profile_images/1434747464/square_400x4...

Given that I only said "saurik" and SD came up with something that not only looks more than quite a bit like me but is more than quite a bit similar to the pose of my profile photo, I'd say clearly that photo would be one of the most important photographs in the database to show up when asking "which human-made source images were used by AI to generate this image"...

...and yet, whatever algorithm is being used here--which I'm guessing is merely "(full) images similar to this (full) image" as opposed to "images that were used to make this image"--isn't finding it; which, to me, means this website is adding more noise than signal to the discussion (in that I think people might learn the wrong lessons or draw the wrong conclusions).