Currently in CC4, a character's eyeball geometry is driven by its skeleton joint. In the Facial Profile, the eight 'Eye Look' expressions (look Up, Down, In, Out) are driven by a combination of bone transforms and mesh deformations for the entire eye area (eyeball, eyelid, eyelash, surrounding facial mesh).

However, ARKit face capture only looks at blendshape deformation; it doesn't know about skeleton rigs. While this works fine within CC4 and iClone, exporting the character for use in a third party program fails, because the eyeballs' rotational motion (joint driven) is not converted, while the surrounding mesh expression (eyelid, lash, etc) continues to work when the blendshape is activated.

What I'd love to see is a simple checkbox at export which would unify the Eye Look expressions and convert the eyeball's rotational data into morph data, combined within their respective blendshapes.
This would work similar to your 'Mouth Open as Morph' export checkbox that already exists, which serves to unify the bone & mesh deformations of the jaw, tongue & teeth into a single blendshape.

This is crucial for ARKit retargeting, as eye tracking is the second most important aspect of face capture besides mouth/lip sync.

Alternately, you could include a character template which doesn't have eye joints at all, whose eyeballs are driven entirely by mesh deforms, in order to work directly with the ARKit model.