Pyramid　selling　over　the　Internet　is　a　new　form　of　pyramid　scheme　with　the　characteristics　of　high　concealment,　strong　deception,　wide　involvement　and　fast　dissemination.　Internet　pyramid　can　disrupted　the　eunomie　of　the　Internet　and　undermine　the　healthy　development　of　the　market　economy　as　well　as　social　harmony　and　stability.　This　study　aims　at　detecting　Internet　pyramid　selling　organizations　by　analyzing　the　contents　of　articles　about　pyramid　selling.　In　the　paper,　we　characterize　the　text　of　pyramid　selling　which　reveals　that　the　introduction　of　investment　returns　(IIR)　can　represent　the　main　characteristics　of　documents　on　pyramid　selling.　We　then　propose　a　new　text　presentation　method　for　text　classification　to　detect　pyramid　selling　documents.　Experiment　conducted　shows　that　the　proposed　method　can　achieve　a　good　recognition　rate　in　the　analysis　of　documents　on　pyramid　selling.