With　the　rapid　development　of　deep　convolutional　networks,　attributed　graph　clustering　has　become　an　increasingly　important　and　challenging　research　area.　In　the　field　of　graph　clustering,　more　and　more　researchers　have　recognized　the　role　of　content　information　and　structural　information　in　the　clustering　process.　However,　existing　methods　often　overlook　potential　issues　that　may　arise　during　the　fusion　of　content　and　structural　information,　such　as　incomplete　structural　information　or　missing　content　information.　These　issues　have　led　to　bottlenecks　in　clustering　performance,　making　improvements　challenging.　Furthermore,　most　of　the　methods　focus　on　minimizing　the　reconstruction　error　of　the　graph　structure,　ignoring　the　embedding　distribution　of　deep　representations,　which　may　result　in　inferior　representations.　To　address　these　challenges,　in　this　paper,　we　innovatively　introduce　an　adversarial　regularized　deep　embedding　clustering　method　based　on　dual　interative-fusion　and　joint　self-supervised　networks,　called　AIJSS.　Specifically,　we　utilize　interative　fusion　techniques　to　deeply　integrate　structural　and　content　information　across　network　layers.　Moreover,　a　triple　self-supervised　module　is　introduced　for　joint　optimization　to　achieve　consistent　and　superior　embedding　representations.　We　also　design　an　adversarial　graph　embedding　module　to　learn　effective　embedding　representation　to　enhance　the　robustness　of　clustering.　Extensive　experiments　on　six　benchmark　datasets　demonstrate　that　our　proposed　AIJSS　method　achieves　significant　improvements　over　current　state-ofthe-art　methods,　highlighting　its　innovation　and　forward-looking　nature　in　the　field　of　deep　graph　clustering.　Our　code　is　publicly　available　at　https://github.com/sliboo/AIJSS.　.