Three episodes for the writer or marketer who's been using AI for drafts and noticing the output reads as 'AI-written' even after editing. The structural reasons AI prose has the patterns it does (it's pattern-matching against training data, not generating from intent), the editing moves that actually change the texture (versus the cosmetic changes most people make), and the workflow patterns where AI helps and the ones where it actively hurts. Built for the writer ready to use AI well rather than either reject it entirely or accept its default output.

The book on editing AI-generated copy into work that sounds like a person wrote it. Covers the structural moves: the pattern recognition for the specific tells AI prose has (the over-balanced sentence rhythm, the conjunction patterns, the abstract noun preference), the editing protocol that breaks those patterns at the structural level rather than just word-swapping, the voice-injection moves that put your specific perspective into the text, and the workflow patterns that prevent AI from being a slop generator. Built for the writer using AI weekly and tired of the output reading as machine-written.

AI-generated copy has tells: the same opening structures, the same sentence rhythms, the same vague intensifiers. Most operators publish AI drafts without running the editing pass that strips those tells, and the audience can read the difference even if they can’t name it. This checklist runs the per-draft editing pass: the opening-pattern audit (catches the "In today’s fast-paced world..." starts), the sentence-rhythm check that breaks up the AI’s preferred long-medium-long pattern, the vague-intensifier sweep ("very," "really," "quite," "leverage," "robust") that strips the words the model defaults to, the personality-injection pass that adds the specific voice the AI flattened, the sensory-detail check that grounds abstract claims, and the read-aloud test that catches what the eye misses. Run it on every AI draft before publish. Pair with the prompt pack for the upstream prompt work; this checklist is the post-draft humanizing pass.

Most AI writing fails at the prompt, not the model. The operator types a vague request, gets bland output, then spends twenty minutes editing what could have been specified upfront. This guide rebuilds the front end of the workflow: the structure of a prompt that delivers an 80% draft, the four context blocks (audience, voice samples, format, do-not-list) that collapse rewrite cycles, the editing pass that catches the AI tells everyone else misses, and the lightweight QA loop that decides what to ship versus what to redo. Includes the prompt-stack that turns a single instruction into a layered brief and the context-injection patterns that get past generic phrasing. Use it when the goal is finished copy in one pass, not draft seventeen of the same blog post.

AI doesn’t sound robotic because the model is bad. It sounds robotic because the prompt asked for "a blog post about X" and the model averaged the entire internet. This listicle catalogs thirteen specific tweaks that swap that average for a particular voice: the "write the way I just wrote" move, the cadence-mirroring trick, the contraction rule, the word-bank list that bans LinkedIn-isms, and the temperature shifts that stop the model from defaulting to consultant-speak. Each entry is one prompt edit, with a before-and-after sample short enough to scan in under thirty seconds. Use it as a desk reference when ChatGPT, Claude, or any model keeps returning copy that reads like a press release. The fixes are surgical, not a full rewrite of the workflow.

Writers split into two camps right now: the AI-skeptics who lose hours rewriting from scratch, and the AI-maximalists whose copy reads like everyone else’s AI copy. The hybrid writer is the third position, and seven days is enough to install the muscle memory. Day one rewires the planning step so the human picks the angle before the model touches it; days two through four cover drafting, voicing, and editing as separate passes with separate prompts; day five builds the personal voice file the model reads back into every job; day six handles publish-ready cleanup; day seven sets the weekly rhythm. Delivered as drip emails, not a binge dump. The output is a writer who ships faster without sounding like a chatbot.

Most prompt packs are a thousand variations of "write a blog post about." The hybrid writer needs the opposite: a tight set of prompts that handle the four jobs that actually take time, with enough specificity that the output lands. This pack covers the draft-to-human conversion (turning generic AI prose into copy that sounds like the writer wrote it), the voice-consistency prompt set (one for short-form, one for long-form, one for replies), the structural prompts that fix flabby paragraphs without flattening them, and the storytelling injectors that add concrete detail instead of stock metaphors. Built for the working session, not the prompt library. Drop them into Claude, ChatGPT, or whatever model is open, and skip the warm-up round of generic output.