Several subgenera have been proposed for splitting out of this massive group. By and by, it is becoming clear how the genus might be split up, with a range of DNA sequence data in large parts agreeing with S.D. Blum's landmark 1988 phylogenetic assessment of osteology
Basically, a core group around the type species Chaetodon capistratus would remain in Chaetodon, while maybe 4 clades would be split off. These could use the names Lepidochaetodon, Megaprotodon and Rabdophorus, and there is one unnamed group containing the Three-banded Butterflyfish (C. robustus) and its relatives. But the monophyly of the Lepidochaetodon group is not fully established; it is both unclear whether Rhombochaetodon is a lineage distinct from Exornator, and whether Lepidochaetodon is indeed closer to these than to any other Chaetodon, particularly to some lineages otherwise placed in Megaprotodon.
Prognathodes, for some time contained in Chaetodon, is worthy of recognition as full genus in any case. Genus status of Parachaetodon and Roa is possible, but not well-studied yet.
The oldest fossils that are usually assigned to Chaetodon date from the late Oligocene, about 25 million years ago. But since it is not easy to distinguish this genus from close relatives, it may be that the Oligocene fossils are actually of other Chaetodontidae. The fossil record of this family is scant indeed, only Chelmon (or some similar genus like Chelmonops or Coradion) being known from Miocene remains. As even crude molecular clocks suggest that the ancestors of the Chaetodon and the Chelmops lineage diverged in the Late Eocene already, nothing more can be said without new fossils being discovered.
Historically, more distantly related fish were placed in Chaetodon too, for resembling them in details – e.g. the common scat Scatophagus argus, which has a similar shape and size, as well as armored larvae like the Chaetodontidae – or simply because they are colorful, smallish, and unusually-looking – e.g. the quite unrelated paradise fish, as C. chinensis.