From time to time one encounters the erroneous rumor that Twitter bots can't reply. Let's take a look at a prolific counterexample. Meet @Guds_barn ("God's children" in Swedish).
Here's @Guds_barn's tweet schedule. This account is quite high-volume - it's likely that the only reason it doesn't always run 24/7 is that it reaches the daily limit of 2400 tweets in fewer than 24 hours.
As seen in the previous tweet, @Guds_barn has used an admirably abundant variety of apps/services to automate tweets. The heavy automated activity starts in early August 2018 - the tweets prior to that are mostly via the Twitter website.
The tool @Guds_barn seems to have settled on for automating tweets is twty. This isn't strictly an automation service; rather, it's an open source project for interfacing with the Twitter API. One could just easily create analytical tools with it as build a bot.
Many of @Guds_barn's tweets are present in multiple languages (mostly English and Swedish, but there are less frequent ones as well).
Let's get back to the topic of replies. 52312 of 61648 @Guds_barn's original tweets are replies, most of which are quote tweets of its own tweets injected into conversations @Guds_barn wasn't previously part of. Is there a pattern in the choice of tweets to reply to?
The answer is yes - each reply tweet seems to have an associated set of terms that trigger the reply. For example, the abortion tweet with the demon emoji is used as a reply to tweets containing "kill", "murder", and/or "abort".
(If you're trying to trigger a reply from @Guds_barn, note that due to Twitter's volume limits you may or may not get one. There are a lot of tweets matching the search terms this account uses.)