For context, consider that for long-exposure astrophotography we are talking about the use of a motorized mount, with at a minimum a drive system on the RA axis of the mount.
The drive system moves the RA axis of the mount at the sidereal rate, allowing the mounted OTA/camera/imager to track the apparent movement of the stars.
All mounts have some degree of tracking error, the sources of which come from minute errors in polar alignment, periodic error in the drive system, mechanical tolerances in various components of the mount, balance issues, and so on.
In the "old days", during long exposure astrophotography, one would either guide the mount using a finder with a specialized reticle, or an OAG and guiding eyepiece, using the fine motion control on the mount's RA axis - manually keeping a guide star centered in the guiding eyepiece's cross-hairs during exposure. This manual correction compensated for the error in the mount's drive system. Depending upon the person, some guided during exposure, while the camera's shutter was open, or provided manual guide correction between exposures.
With the advent of the SBIG ST-4 camera, it was possible to use TTL-level signals to control a specially wired mount to replace the manual guiding with auto-guiding, where a CCD camera with fine pixel accuracy and corresponding software on a host computer sent corrective guide signals to the mount, emulating the button-presses of a hand controller to move the mount in both DEC and RA.
This has been refined over the years, such that commodity mounts now have embedded ST-4 "guide ports" that conform to the defacto standard set by SBIG vis-a-vis the ST-4 electrical and pulse protocol.
The guider provides continuous correction during exposure (or more accurately independent of/decoupled from exposure).
The guide correction is both tunable, and minimal, and does not result in blurring of the image during the exposure - in theory, the guide correction keeps the target object "centered" over the same pixel group on the imaging array (or film).
The same thing holds true for use of an OAG while manually guiding. The motion required is slight, continuous (usually), and keeps the target object centered over the same pixel group on the imager by using the crosshairs of the eyepiece mounted on the OAG's barrel. No blur (if you're doing it right).
CGEM 800 HD, NexGuide, Orion XT8 Limited Edition, Oberwerk BT-100, Canon 20D/20Da/T3i/60D/5D Mk III, various eyepieces, adapters, geegaws, widgets, and tiddlybits