[aaronblohowiak]

It is a little more nuanced. Twitter is claiming than fewer than 5% of the monthly active users they report are spam bot, not that fewer than 5% or those on the platform. If they “don’t count” them but also don’t ban them, that would leave room for bit theories to be “true”

[Majromax]

There's also a deeper set of definitions at play. Twitter the corporation cares about monetizable active users, as stated. Or, in English, "what fraction of ads served are really going to robots?"

Twitter users don't give a fig about Twitter's ad impressions. They care about what fraction of comments or interactions are in fact with bots or astroturfers. From the perspective of Twitter as a communications medium rather than an advertising channel, this is the relevant statistic.