The early days of first person shooters like Quake and Unreal Tournament gave birth to one of the most unique types of video game AIs: the FPS bot. Within a few years time these mindless enemies became an institution. Then just as quickly as they arrived, they disappeared. For many in the late '90s and early 2000s, the ability to run through maps blasting things with a rocket launcher predated their access to high-speed internet. This created a need for offline opponents that could understand the game rules, collect powerups and provide a semblence of a challenge to the modemless.
Quake's Reaper Bot might be the first example of the trend, according to Super Bunnyhop. Rather than learn its environment, Reaper was programmed to navigate the map by following the movements of human players, while shooting at them with near-perfect accuracy. The result was an AI that was difficult to beat but also easy to confuse. Standing still was enough to trip up Reaper.
Bots only became smarter and more ubiquitous from there, learning to imitate normal gameplay by following invisible waypoints or chasing after items. Early console shooters like Perfect Dark adopted bot players because the Nintendo 64 didn't have internet connectivity, and many more games, like TimeSplitters, followed suit.
Improved broadband internet access started to make legitimate multiplayer gaming a real threat to bots, but the death blow came from Moore's Law. Faster computers created the opportunity for bigger, more complicated maps. The more open-world these maps were, the harder it was to program bots that could simulate some level of gameplay fluency. Games like Halo came along to prove, beyond all doubt, that the age of the bot had ended.