Why do you want to know if a machine or human wrote a tweet? What difference does it make, and where do you draw the line? If I use Grammarly to correct my language in a tweet then did a machine write it? How about if I generate text using GPT-3 and then manually edit it?
One is more labour intensive and more expensive. The other can be done at scale for cheap. The negative impacts most likely don’t scale linearly, so volume is an outsized concern.
Please define your terms.