Depends on the platform. On Reddit, definitely yes, as almost all censorship is "deletion style" where the target post or comment is initially visible until the mods suppress it.

But your point is well taken, it's definitely more challenging and expensive than what's being done here.

It would be one heck of a challenge, but, almost by definition "impossible" - as a naive guess, perhaps (for those platforms that have opened code) characterize some statistical signatures that "arbitrary data" ran through it (simulating user activity) would show ...

... and then check for those (or, counterexamples) in the wild.-