pythonation/ai-text-humanizer-protocol

This repository, called Humanizer Protocol v2.5.1, is not a program you install. It is a long, detailed instruction text known as a system prompt, written for chatbots like ChatGPT and Claude. You paste it in as the first message of a conversation, and from then on the chatbot is supposed to behave like a strict editor whose job is to rewrite AI-flavored writing into prose that sounds like a real person wrote it. The README is mostly in Arabic, with the actual prompt body in English. It explains that the protocol draws on the Wikipedia community's WikiProject AI Cleanup, which catalogues the telltale habits of AI-generated text. The four-step workflow described in the README is to scan the text for overused phrases, cut padding and machine-style apologies, inject human rhythm and specificity, and then run a final anti-AI audit pass before returning the result. The prompt itself is broken into named sections. There is a content section that bans things like puffed-up significance statements, vague references to experts or industry reports, and formulaic challenges-and-future-prospects paragraphs. There is a language section that lists banned words such as delve, tapestry, crucial, pivotal, and landscape, and rules against forced groups of three, synonym cycling, and the not just X but Y construction. A style section attacks the visible fingerprints of AI text. It removes em dashes, mechanical bold formatting, title case headings, emojis, and curly quotation marks. A communication section strips chatbot pleasantries like Certainly, Of course, and I hope this helps, plus the disclaimers AI models often add about knowledge cutoffs. The prompt also has a voice calibration mode. If you give it a sample of your own writing, it is told to analyze sentence length, word choice, and punctuation habits in that sample and mimic them. If you do not give a sample, it falls back to a default voice that values opinions, varied rhythm, mixed feelings, first-person honesty, and specific concrete details over neutral polished summaries.