Run in order. Prompts 1-4 are the positioning work — the most leverage and the most rushed. Don’t move to prompt 5 until prompt 4 has produced a line a competitor would have to lie to copy. If it could be cut and pasted onto a competitor’s site without anyone noticing, you don’t have positioning yet — you have a slogan.

1. The claim only you can make. “Here is my MVP from M5, my model from M3, and my segment from M2: [paste]. Draft 10 candidate positioning lines, each one sentence, each claiming a specific thing about what I do, who it’s for, and why no-one else does it the same way. For each, list the 3 competitors most likely to push back, and for each competitor, write the literal sentence they would have to say to claim the same thing. Mark any line where one or more competitors could honestly say the same sentence — those are eliminated. Output only the surviving lines.”
2. The lie test. “Take the surviving lines from prompt 1. For each, write the sentence a direct competitor would have to put on their own site to make the same claim. Then label each competitor sentence: (a) true and they already say it, (b) true but they don’t say it, (c) would be a lie if they said it. Only lines where every competitor falls into category (c) are owned. Anything in (a) or (b) is table stakes dressed up as positioning. Output the owned lines only.”
3. Earn the claim. “For each owned line from prompt 2, write the proof stack — the 3 concrete things in my background, my product, my method or my track record that earn the right to say this. Proof is specific: a named project, a number, a year, a partnership, a result. If I can’t list 3 specific proofs, the line is aspirational, not owned — flag it. Rank the remaining lines by strength of proof. The top line is the positioning candidate.”
4. The one-line positioning lock. “Refine the top line until it does five things at once: (a) names what I do in plain English a buyer would use, (b) names who it’s for specifically enough to exclude people, (c) names the thing only I can credibly claim, (d) reads in under 12 seconds, (e) survives being said out loud to a sceptical peer without sounding like marketing. Iterate the line at least 5 times. Output the final line, plus the one sentence I say next when someone asks ‘what does that actually mean’.”
5. Voice rulebook — what I sound like. “Based on the positioning line and on 3 samples of how I actually write (paste samples below: an email, a long post, a message to a friend), extract my voice rules in plain language. Cover: (a) sentence length tendency, (b) words I reach for that most people don’t, (c) words I refuse to use, (d) how I open, (e) how I close, (f) what I do with jokes / metaphors / numbers, (g) the energy register — warm / sharp / dry / formal. Output as a one-page rulebook that another human (or an AI) could follow to write in my voice without me in the room.”
6. The ‘never say this’ list. “From the voice rulebook and the positioning line, write the explicit ban list — 15-25 words, phrases, and framings I will never use in brand copy, even when the AI suggests them. Cover category-default platitudes (‘innovative’, ‘unlock’, ‘game-changer’), genre-fatigue phrases that are everywhere in my space, and any sentence shape I find on three competitor sites. For each ban, write the one-line reason. This is the list I paste into every Claude / GPT writing prompt as a constraint.”
7. Visual system — the discipline, not the design. “My positioning is [paste]. My voice is [paste rulebook]. Recommend a minimum viable visual system: (a) primary typeface and pairing — name 2 specific font choices and why, (b) colour palette — 1 primary, 1 supporting, 1 accent, with hex values and the reasoning behind each, (c) layout rule — the one structural principle that runs across every page / post / asset (e.g. ‘always left-aligned, always one image per fold, always one CTA per page’), (d) photographic / imagery approach — what subjects, what crop, what light, what’s banned, (e) the one design move I borrow from no-one else in my category. Bias to fewer choices. The system fits on one page.”
8. The touchpoint audit. “List every customer touchpoint in my business today — website pages, product packaging or screens, email templates, social handles, invoices, voicemail, business card if any. For each, ask: does this touchpoint carry the positioning line, follow the voice rulebook, and apply the visual system? Output a table: touchpoint, status (on-brand / drift / nothing), the one specific fix that would bring it on-brand, and the time to fix in hours. Sort by leverage — touchpoint the most buyers see first goes first.”
9. The discount-killer check. “Imagine a buyer is comparing me side-by-side with 3 named competitors at similar price (paste competitor names). Based on my positioning, voice and visual system, what is the buyer paying for that justifies my price holding when theirs drops? Write the exact answer a buyer would give a colleague who asked ‘why did you pick them and not the cheaper one’. If I can’t write that sentence, my brand is currently negotiable on price — flag the gap and tell me which of the three pillars (positioning, voice, visual) is the weakest link.”
10. The brand brief. “Write my one-page brand brief, suitable to print and pin next to the MVP brief from M5. Sections: (1) positioning line, (2) the proof stack behind it, (3) voice rulebook in 5 bullets, (4) never-say list in 10 items, (5) visual system — type, colour, layout, image, signature move, (6) the top 5 touchpoints and their current status, (7) the discount-killer sentence. Plain English. No jargon. Fits on one A4 page. The page anyone on my team — or any AI I prompt — reads before producing anything in my name.”