So ChatGPT 5.4 finally arrived in my app and I put it to the test with a current project that I’m working on. The initial output was already pretty good but I still had to guide it along with my specific requirements and writing style. After ten minutes of prompting and refining, I thought to myself: “I don’t think anyone can tell this was written by a machine.” so I pasted the output into four different AI detectors.
Indeed! All four of them – GPTZero, Grammarly, ZeroGPT and Quillbot – declared that this was largely written by a human. By now, we know that AI detectors aren’t 100% reliable, but GPT-5.4 (and probably the other new models) has totally decimated their business model.
But I’m not worried, because I am still holding the reins here and I managed the entire conversation from beginning to end. The value of the data that I provided and my ability to shape the project goals were critical to producing the final output. So like I always say, in the age of AI, everyone can “write” well, but not everyone can think and direct AI well.