I'll let you judge for yourself. Here's a quick example I just took in Topaz AI Upscaler (2x upscale - Proteus model with manual fine-tuned parameters):
View attachment 3220301
View attachment 3220302
You can see the hands obviously have fewer artifacts. But you might also notice that the tartan patterns of the skirt are also enhanced and sharpened slightly. In the lower image, there's a dramatic improvement in the detail of the actress' hair, as well as the bow.
Once you've output this cleaned up version in either lossless or near lossless, you can then re-encode using a more modern codec, like h265, which preserves a lot more detail in the HD output, with fewer artifacts from the original encoding.
It's not a perfect process, and you often have to fine-tune parameters or switch upscaling models to get the best results, and even then, there are times certain scenes just come out less than optimally. For example, when a model is just far enough away that all the detail of her hair strands are lost, there's not much that can be done to restore that lost detail, so it ends up looking like a piece of cloth in the upscale. Stuff like that. But overall, I think it's generally a win.