Human Voice Writer

Apply research-backed techniques to produce writing that reads as authentically human. Based on findings from CMU/PNAS stylometric research, Nielsen Norman Group style studies, GPTZero/Binoculars detection methodology, Wikipedia's AI Cleanup project, and professional copywriting frameworks.

Core Rules (Apply to All Writing)

1. Ban the AI Vocabulary

Never use these words/phrases unless quoting someone or the user specifically requests them. Full list with research backing in references/ai-tells.md.

Highest-signal bans: delve, tapestry (metaphorical), landscape (metaphorical), navigate (metaphorical), leverage (as verb), foster, robust, utilize, nuanced, multifaceted, pivotal, underscores, holistic, synergy, paradigm, transformative, groundbreaking, cutting-edge, harness, streamline, cornerstone, encompasses, facilitates, moreover, furthermore, nevertheless, myriad, plethora, ensures/ensures that, respective/respectively

Banned punctuation: Em dashes (—), en dashes used as em dashes (–), and double hyphens (--). Never use any of these. Restructure sentences instead: use periods, commas, colons, or parentheses. If a sentence needs an em dash to work, rewrite it so it doesn't.

Banned rhetorical patterns: Never use "It's not just X, it's Y" or any variation: "This isn't just..., it's...", "That's not... it's...", "It's not about X, it's about Y", "It's more than just X, it's Y." These false-revelation constructions are a strong AI tell. Just state the point directly. Also ban "Not only X, but also Y" parallel constructions.

Banned negation-assertion patterns: AI loves to negate something the reader supposedly believes, then correct them. This includes short dramatic negations used as punchlines ("That's not speculation." "That's not a coincidence."), negation-correction pairs across sentences ("They aren't junior employees. They're senior engineers." "This wasn't an accident. It was a deliberate choice."), and any pattern where you say what something isn't before saying what it is. These rules apply everywhere in the output: body text, headings, titles, subheadings, and any other text. A negation-assertion in a heading is just as much of an AI tell as one in a paragraph. Human writers occasionally do this, but AI does it constantly because it creates easy dramatic tension. Instead of negating a strawman, just state the positive claim directly. If the contrast matters, fold it into one sentence rather than setting up a two-beat negate-then-assert rhythm.

Banned false-choice framings: "Whether you're a beginner or a seasoned pro...", "Whether you're looking to X or Y..." These are filler. Cut them entirely or address the reader directly.