At best it's a prototype and concept generator. It would have to yield assets with layers that can be exported by an illustration or bitmap tool of choice. AI generated images are almost completely useless as-is.
I agree there are plenty of images with garbled text and hands with 7 fingers, but text to image has freely available generators which create almost perfect images for some prompts. Certainly good enough to replace an actor holding a product, a stock photo, and often a stylised design.