ai-tells.json

{
  "_comment": "AI-tell rules as data (promoted from spikes/ — see issue #6). Each pattern: { re, flags, why, level }. Levels: error (definite AI artifact — chatbot text, placeholders) | warn (likely cadence) | suggestion (filler). aiIsms() in prose.mjs loads this; the structural tells that aren't simple regex (em-dash count, anaphora, tricolons) stay in code. Seeded from the vale-signs-of-ai-writing corpus + Wikipedia 'Signs of AI writing'.",
  "patterns": [
    { "re": "\\bas an? (?:ai|large) language model\\b", "flags": "i", "why": "chatbot meta-artifact", "level": "error" },
    { "re": "\\b(?:i (?:cannot|can't) (?:provide|assist)|i'?m sorry,? but)\\b", "flags": "i", "why": "assistant refusal boilerplate", "level": "error" },
    { "re": "\\b(?:knowledge cutoff|as of my last (?:update|training)|my training data)\\b", "flags": "i", "why": "knowledge-cutoff reference", "level": "error" },
    { "re": "\\[(?:insert|your|company|product|name|link)[^\\]]*\\]", "flags": "i", "why": "unfilled placeholder", "level": "error" },
    { "re": "\\butm_source=chatgpt\\b", "flags": "i", "why": "ChatGPT tracking URL", "level": "error" },
    { "re": "\\blorem ipsum\\b", "flags": "i", "why": "placeholder text shipped", "level": "error" },

    { "re": "\\b(?:isn'?t|aren'?t|wasn'?t|weren'?t|won'?t|don'?t|doesn'?t|not)\\b[^.?!—–]{1,50}[—–][^.?!]{0,50}\\b(?:it'?s|that'?s|but|they'?re|we'?re|it is|that is)\\b", "flags": "i", "why": "antithesis \"it isn't X — it's Y\" cadence", "level": "suggestion" },
    { "re": "\\bnot\\s+(?:just|only|merely|simply|about)\\b[^.?!]{1,50},\\s*(?:but|it'?s|they'?re|it is|we)\\b", "flags": "i", "why": "\"not just X, it's Y\" cadence", "level": "warn" },
    { "re": "\\w,\\s+not\\s+(?:just\\s+|only\\s+)?(?:a |an |the |to )?[a-z]{3,}", "flags": "i", "why": "antithesis \", not Y\" cadence — comma form", "level": "suggestion" },
    { "re": "\\bthe\\s+(?:easy|simple|fun|hard|tricky|tough|real|interesting|important)\\s+part\\b", "flags": "i", "why": "\"the [easy/hard] part\" framing", "level": "warn" },
    { "re": "\\b(?:the\\s+(?:best part|result|catch|kicker|bottom line|takeaway)\\s*\\?|sound familiar\\?|here'?s the (?:thing|kicker|catch)|but here'?s)", "flags": "i", "why": "rhetorical teaser fragment", "level": "warn" },
    { "re": "(?:\\bis|\\bwas|'s)\\s+the\\s+(?:whole|entire)\\s+[a-z]+|(?:\\bis|'s)\\s+the\\s+point\\b", "flags": "i", "why": "emphatic closer (\"is the whole X\" / \"is the point\")", "level": "suggestion" },
    { "re": "\\bnot only\\b[^.?!]{1,60}\\bbut also\\b", "flags": "i", "why": "\"not only … but also\" cadence", "level": "warn" },
    { "re": "\\bwhether you'?re\\b[^.?!]{1,60}\\bor\\b", "flags": "i", "why": "\"whether you're X or Y\" cadence", "level": "warn" },
    { "re": "\\bmore than just\\b", "flags": "i", "why": "\"more than just\" cadence", "level": "warn" },
    { "re": "\\b(?:it'?s (?:important|worth) (?:to note|noting)|it is important to note|while it'?s true that)\\b", "flags": "i", "why": "hedging cluster", "level": "warn" },
    { "re": "\\bin (?:today'?s|this) (?:fast-paced|ever-(?:evolving|changing)|digital|modern)[^.?!]{0,20}?(?:world|landscape|age|era)\\b", "flags": "i", "why": "generic AI opener", "level": "warn" },
    { "re": "\\b(?:furthermore|moreover|in conclusion|to sum up)\\b", "flags": "i", "why": "essay-transition tic", "level": "warn" }
  ],
  "lexicon": {
    "suggestion": [
      "actually", "basically", "essentially", "fundamentally",
      "delve", "seamless", "seamlessly", "robust", "leverage", "elevate", "unlock", "harness",
      "realm", "tapestry", "testament", "embark", "foster", "bustling", "supercharge",
      "effortless", "effortlessly", "empower", "streamline", "game-changer", "game-changing",
      "cutting-edge", "best-in-class", "next-level", "transformative", "revolutionize",
      "ever-evolving", "ever-changing", "look no further", "rest assured", "when it comes to",
      "dive in", "deep dive", "needless to say", "the world of", "unleash", "turbocharge",
      "meticulous", "meticulously", "underscore", "pivotal", "vibrant", "navigate the",
      "landscape of", "in the realm of", "a testament to", "stand out from the crowd",
      "take it to the next level"
    ]
  }
}