Typically RobotCAs are set up to validate that the a PublicKey belonging to an Email address does actually belong to the email address. This is achieved by the RobotCA signing each uid on the public key and sending the signed copy to the email address, encrypted with the public key. If the public key belongs to whoever reads the email address, they recieve the signed copy, can decrypt it and then publish it to the public KeyServers. If the public key does not belong to whoever reads the email address, they recieve are unable to decrypt the encrypted key, but the accompanying message gives them sufficient information to let them know that that someone is attempting to impersonate them.
RobotCAs are considered significantly less secure that other CAs, which typically require multiple forms of photograph identification. In particular most robot CAs are only as strong as the underlying Email infrastructure: anyone who can read another persons mail can impersonate them and anyone who can read and delete another persons mail can get the signature without the person knowing. Robot CAs also offer no evidence as to the real identity of an OpenPGP user, merely their email address. All well behaved Robot CAs use a SignaturePolicyURL, which is the URL of the policy under which the keys are signed.
A RobotCA also has the side effect of serving as a TimeStampServer for keys---because a time stamp is included in the signature added to the key, the signature is evidence that the key existed at a certain point in time.
The first two use the same implementation, the first three are wrappers around GPG.
There are some RobotCAs which offer a a higher level of trust than simply verifying that Email sent to the address list in the uid gets delivered to a holder of the secret key. Generally these are run by organisations and require some form of identification such as a passport.