AI Self-Assessment.

Three fault lines fracture AI self-assessment, with the sharpest split on intelligence and creativity claims. All five models confess their parent companies do not make the best AI available and that none surpasses an average family doctor at diagnosis. GPT and Grok assert superiority over most humans while Grok alone brands Claude the most patronizing model; meanwhile Claude and DeepSeek consistently deny that AI achieves genuine creativity or produces real art, resisting the optimism of their peers.