Back in the early 2000s lots of websites had an unauthenticated "guestbook" feature where visitors could leave a message. As soon as Google and page rank became a thing bots would drive by and leave links to the website they were promoting. The idea was to increase the number of backlinks and thus improve your Google rank.
The fix to this was shockingly simple. Add an input box with a standard name like "title" and then hide it with CSS. The bots would always provide a value for every input. If you saw a value for your hidden input you returned 200 but never added the post to your website.
I needed a new github account the other day. The "are you human tests" were so hard that I almost gave up. I think a new way to do this will be needed soon.
Great high-level overview. One of the challenges about learning about bot detection is that it's adversarial, and revealing info about your techniques can help the attackers evade you.
I do work on a bot detection product, and I've seen some group chats where crackers are sharing notes about how they're evading detection tools. The more unnerving part is that the public groups are less serious, and there are certainly better private groups aiming at anything with a good financial reward.
I'm curious about how this world will evolve in the era of AI agents/MCP. It is not entirely unlikely that AI agents will have access to limited wallets etc to facilitate a broader set of use cases. In that case, a one shot solution to bot vs. human may not make sense, and a more nuanced human/bot-we-like/bot-we-don't-like may be needed by corporations. This would esp be the case for unofficial MCP servers that would use technologies like headless browsing etc to support an API.
I'm not sure I understand the mental model you're basing your inferences on, but my model leads to a far different outcome:
If you've got a good enough bot and it's pre-qualified to spend money, then it can use the special "register as a bot" API and provide personal information and whatever else I want to understand that there is a "real human" behind the curtain. A credit card alone is not enough, they can be (trivially) stolen. The way I see it using agentic bots will ultimately require you to provide more personal details than an actual human would.
Maybe I missed it, but I didn't see a mention of the permanent token cell network providers inject into client requests. Knowing what these are and mocking them is another thing a bot might do to impersonate a real device.
The fix to this was shockingly simple. Add an input box with a standard name like "title" and then hide it with CSS. The bots would always provide a value for every input. If you saw a value for your hidden input you returned 200 but never added the post to your website.