I believe Zignal (which is sort of CrowdTangle for Twitter) provides a variable in their API for "likelihood they are a bot" which is direct from Twitter -- so

I believe Zignal (which is sort of CrowdTangle for Twitter) provides a variable in their API for "likelihood they are a bot" which is direct from Twitter -- so Twitter often knows the account is a bot, but much less often takes action.