An algebraic stack is essentially a geometric stack on Cring?: it denotes either a Deligne-Mumford stack or a more general Artin stack? in the traditional setup of algebraic spaces: they are (Grothendieck) -stacks of groupoids on the étale site? satisfying additional representability conditions.
An algebraic stack is required to have an atlas from an algebraic space (or in a stronger version a scheme) ; the morphisms of the cover are required to be
surjective representable and étale (for a Deligne-Mumford stack);
surjective representable and smooth (for an Artin stack?).
Here we used the following definition. A map of 1-stacks on the étale site is a representable morphism of stacks if for any scheme the fiber product is a stack associated to a scheme.
Let be a property stable under composition and pullback. A representable map of stacks is said to have a property if for any scheme and a morphism of 1-stacks the pullback has property .
Urs Schreiber: don’t we also need to demand that the diagonal morphism is a) representable morphism of stacks, b) separated, c) quasi compact ?
See for instance def (4.1) in Laumon, Moret/Bailly Champs algébriques.
Nowdays people talk about differentiable stacks and topological stacks, meaning analogues of Deligne–Mumford (DM) or Artin stacks defined in the setup of the category of manifolds or some convenient category of topological spaces. Orbifolds are an example of an Artin stack. For orbifolds stabilizer groups are finite, while for Artin stacks in general they are algebraic groups.
A noncommutative generalization for Q-categories instead of Grothendieck topologies, hence applicable in noncommutative geometry of Deligne–Mumford and Artin stacks can be found in
For the topological variant, see topological stacks and references therein; and for differentiable stacks