Bots are automated programs that act like IRC clients and maintain a presence on an IRC channel in the absence of an actual human being.

There are many kinds of bots, some of them are very useful, some of them are destructive. They basically act on their owners behalf, and so like their owners, can be beneficial or destructive.

Examples of useful bots include service bots which provide some service such as the dispensing help files in response to user requests. There are bots that provide dice rolls for fantasy role playing games, bots that provide card draws for card games, bots that dispense files of real-time data such as images of the shoemaker-levy impact on Jupiter.

The most common form of a bot is what is known as a channel guard bot. In IRC channels are dynamically created and, when everyone leaves them, destroyed. Many people meet on a given channel regularly. To assure that channel is not taken over by another group, they will run a channel guard bot. Channel guard bots are also used to control content. For example, a channel guard bot can be set up on a channel intended for children to automatically kick anybody out that uses inappropriate language.

Channel guard bots which are the most common type, would not be necessary if servers maintained channel state information rather than discarding it. To do this or not to do this seems to be a bit of a religous issue amoung operators, but from the users perspective it seems that it would be a pretty desirable feature.

There are also destructive bots. For example, clone bots launch many copies of themselves and flood a server with connection requests, disrupting normal traffic. Floodbots flood a channel, generally with worthless ASCII graphics, making that channel unusable, and then there are nick-collide bots used to try to hack operator status on a channel. We do not welcome these kinds of bots.

Bots consume resources just as a human operated client would. If the capacity of the network is inadequate, then this will cause lag, just as human operated clients will cause lag in an overloaded network. One of the problems with a network based on a star topology is that all of the network traffic must pass through some central HUB, and that places an upper limit on the growth potential for the entire network. A more fully meshed topology avoids this problem by allowing traffic to be distributed over a number of paths which grow as the number of nodes grow.