In order to counter this, one thing we’ll need is to look at how likely it is that someone is human. We could of course rely on centralised ID provision, but that has its own problems.
Could we not turn this around and use all that add tracking stuff to build a score on how likely it is someone is real? Browsing history alone isn’t enough, you also need to look at whether they ever bought anything etc. That score could then be used to deprioritise non-humans. Of course, this has flaws, but may be a step in the right direction. If we had an industry consortium working on this, literally for the good of the world? Once it has traction, you could then open the door to offline “proof of being human” — that has always failed in the past because it started there, but if you already have a score, you’d be more inclined to want to improve it?