The matrix of sources for a large picture can be mind boggling huge.
We have jpeg, webp, avif and jpegxl. Portrait orientation and landscape. 3 or 4 different pixel ratios and of course, different screen resolutions. Dark mode support doubles all of that work. Hardly anyone is going to do all of this correctly, automated or not.
I've primarily been using AVIF with JPG as fallback on my blog for a few years. The efficiency gain is wild. I wonder if AV2 will be integrated into AVIF at some point (AVIF is based on AV1).
In other news, AVIF is at 95%: https://caniuse.com/avif so it's probably time to give it the first preference over jpg.