There is no way the bot share is so low among active users? When I look at replies to some celeb tweet there are so many obvious bots.

And yet you've probably never seen my tweets, because I'm generally not replying to celebs' tweets. Bots (owners) are trying harder than most people to maximise their visibility before getting banned, so it's expected that you would notice a higher proportion than they actually represent. That's not to say I have any reason to think the bot share isn't higher, just I don't think it's possible for any single person to judge anywhere near accurately considering the huge scale of Twitter and that therefore any one of us users is only seeing a tiny, tiny fraction of the accounts/tweets being posted.

Maybe consider where you're looking is more likely to attract bot replies? It's like an NFT discord you're more likely to see spam for eventual rugs because they're hunting people already proven to be likely to invest in stupid NFTs. Step off celebrity twitter and there's a lot less bot activity because it's less profitable to spam there.

Conditional probabilities. Celeb replies is the most likely place you'll find bots.

I'm dying to know how they're defining "bots" for this purpose and whether it's specifically defined in the agreement. I suspect a (perhaps feigned) argument over the scope of the definition will be the deal's undoing.