Every prompt is ours
There are excellent open prompt libraries on GitHub — we read them too. But republishing someone else's prompts would make this page a duplicate of a page that already exists. So we wrote our own, from the work we do.
Not 500 prompts scraped from a list. A small, deliberate set — each one written for a real business job, tested on live client projects, and explained so you can rewrite it for your own.
There are excellent open prompt libraries on GitHub — we read them too. But republishing someone else's prompts would make this page a duplicate of a page that already exists. So we wrote our own, from the work we do.
Most prompt libraries are built for developers. This one is built for owners, marketers and ops leads — the people whose Tuesday is proposals, ad copy, review replies and a report that's overdue.
A prompt you can't modify is a dead end — the moment your job differs slightly, it breaks. Each page explains why the prompt is built the way it is, so you can rebuild it for work we've never seen.
Prompt libraries usually sort by "writing / coding / creative". That's a taxonomy for hobbyists. We sort by the function that owns the work and the outcome it's judged on.
Content briefs, competitor gaps, schema, and the one that matters most now — finding out why AI assistants recommend your competitor instead of you.
Turning discovery calls into proposals, handling the objection you always get, and writing the follow-up that doesn't sound like a follow-up.
Ad copy that isn't beige, a month of social from one customer conversation, and email that gets opened by people who've ignored you twice.
Review replies that read like a person wrote them, refund conversations that keep the customer, and an FAQ built from real tickets.
Meeting notes into decisions and owners, an SOP written from watching someone work, and job ads that filter properly.
Reading a messy spreadsheet honestly, a monthly report your board will actually finish, and competitor research that isn't just their homepage.
We publish a prompt only once it has earned its place on real client work. That makes this list grow slowly — deliberately. Two are live now; the rest are in testing.
Find out what ChatGPT, Claude, Gemini and Perplexity actually say about your business — and why they recommend your competitor instead of you. This is the prompt we run in the first hour of every GEO engagement, and the results are usually uncomfortable.
Turn a messy discovery call into a proposal that argues for the client's outcome instead of listing your deliverables. Built to kill the three things that lose proposals: vague scope, unanchored pricing, and a document that reads like it was sent to fifty other people.
Turn any keyword into a ready-to-write brief — titles, meta, a full outline and the questions to answer.
Five click-worthy, correctly-sized title tags and meta descriptions, keyword-optimised and on brand.
From one seed topic to pillars, intent-sorted clusters and a page-by-page plan with quick wins.
Benefit-led page copy — hero, sections, objection handling and FAQ — with your keyword woven in.
Eight answer-first Q&As plus valid FAQPage JSON-LD, built to win featured snippets and AI answers.
Three honest, confident responses to any objection — quick reply, email and call script.
Five subject lines, a tight prospect-first cold email, and a follow-up that earns replies.
Turn one transcript, article or notes into 12–16 platform-ready posts with a 4-week cadence.
Subject lines, a scannable body, a PS and a plain-text version for any nurture, promo or announcement.
On-brand public replies to any review — positive or negative — plus an optional private follow-up.
Empathetic, accurate replies for tickets and chat — with a short version and a de-escalation variant.
Turn raw notes into a summary, decisions, owner-assigned action items and a ready-to-send recap.
Turn a rough description into a full SOP — purpose, roles, numbered steps, checks and pitfalls.
A side-by-side competitor comparison, the gaps you can own, and five prioritised 90-day moves.
Turn raw metrics into a plain-English report — summary, context, wins, a lesson and next steps.
These are running on live client work right now. Each gets published once it survives contact with a real deadline.
TEACH is the framework we teach in AI Essentials — five questions every prompt should answer before you press enter. It's why our prompts are long, specific, and boring to look at. Weak prompts are short because they leave the model to guess; strong prompts remove the guessing.
Learn the framework and you stop needing prompt libraries at all — including this one. That's the point. We'd rather you write your own than bookmark ours.
Context the model can’t infer. It doesn’t know your industry, your customer, or last quarter. Anything you leave out, it invents.
A role with real constraints. “Act as a marketer” does nothing. “You’ve run paid social for trade suppliers for ten years” changes the vocabulary.
The shape of a good answer. Format, length, structure. If you don’t specify it, you get an essay when you needed a table.
An example of “right”. One sample of the output you want beats three paragraphs describing it.
Permission to push back. Tell it to ask before assuming, and it stops confidently inventing what it doesn’t know.
We sell SEO, AEO and GEO — getting businesses found on Google and cited by AI assistants. So we built the library the way we'd build yours: original content, structured for machines, aimed at people who buy. If you found this page through an AI assistant, that's the whole argument.
Talk to us about your visibility