I work in bot protection and it's a sound idea but doesn't really work in practice. As long as there's more than 1$ of value to be gained it's worth it for the bot makers.
This also makes it so that botting is only accesible to select few actors that have the required resources i.e. russian troll farms or large bot networks from china, in turn this increases their value. This is very good for them.
Reality is that the only way to stop bots is to constantly change up the detection system. This is called a "cat and mouse" sort of problem and it really is the only way to do it. The attacker always has to catch up and it can be trivial that takes them couple of hours to do but it also reveals behavior patterns for marking bot accounts. This actually works really well in practice but requires a lot of dev resources and many companies low-key like bots which is another thread entirely.