Afaik, nothing except the most advanced bots could overcome the bot-tests I proposed. Im talking NSA/espionage level heuristic AI bots involving very complex integration of multiple softwares, and entire teams of programmers. Current EVE bots are nowhere near that capacity, nor the capacity of EVE botters to acquire or know how to use such complex bot systems.
Furthermore, if the bot-tests are randomly selected from a pool of several, we reduce that margin even further, as the bot would have to be capable of dealing with any number of very different bot-tests, as the EVE UI/client allows for.
CCP can periodically issue new forms/variants of my suggested bot-tests, that will wreck existing bots until reprogrammed to deal with the new ones, IF they can even break the existing ones, which I think they cant.
This puts CCP one step ahead of bots, always.
A human, however, can resolve them in a matter of maybe 5s and 1-5 clicks.
The ideal, ofc, is as fast a bot-test with as few clicks, for greatest efficiency vs bots, so as to not burden non-bot players unduly.
For casual readers, this is not about a “captcha” with text, numbers or images you are used to when browsing the web. Im talking about bot/AI- tests that use the capacity of the EVE UI/client in ways a web browser cannot.
For you, it will just look like a very simple minigame that you can solve easily in less than 5s with very few clicks.
For bots, it will fk their ■■■■ up.
They cant deal with these.
To you its simple.
To bots its like an ant trying to cure cancer.
Something perhaps like the following examples:
What about a moving element to click in a pop-up?
In the below image, the center of the vortex moves and you click it with the reticle cursor:

Or a simple sliding bar puzzle?
In the below image, you move the bars up and down to reveal the 1/2/3:

In the image below, you must rotate each disc so the numbers align between the green lines:

Good luck to any programmer trying to code a bot for the disc example at the bottom. It will have to identify the randomized number locations, and turn the disc for each to match inbetween the green lines. Plus CCP can program the test so the green lines are randomly placed on top/bottom/left/right of the grid, so the bot also has to recognize into which space it has to turn all 3 discs so the numbers all are within it.
Sure, perhaps someone can do that after expending a great deal of time/effort. But what if the next pop-up bot test isnt this one as randomly selected from the test pool? What if CCP changes the dimensions of the discs, colors, numbers to images etc in the next update?