> It's definitely a rough estimate but about ~65% of text content was determined to be a bot.
A scary number. I wonder about a per-subreddit distribution, though. I imagine the primary subreddits have slightly worse human-to-bot ratio, niche subreddit somewhat better, with non-political, non-easily-monetizable subreddits having the best.
Did your analysis also attempt to identify troll farms? Would the content produced by protein bots be grouped in the ~65% of bot content, or the remaining 35%?