The gap is not between people who use AI and people who do not. It is between people who know what good copy is and people who do not.
Most people using AI for copy are not getting bad results because the tool is bad. They are getting bad results because they do not know what good looks like.
Here is what happens. You open ChatGPT. You type something like: "Write a caption for my skincare brand launching a new moisturiser." AI gives you something back. It is grammatically correct. It has a structure. It might even have a hook. You post it. Nothing happens.
The problem is not the tool. The problem is that you handed AI a task without giving it the information a copywriter would have before starting. And because the output looked like copy, you could not see what it was missing.
AI can produce. It cannot decide. Those are two different things. A production line can produce thousands of chairs a day. But someone had to decide what type of chair should be produced, and what the standard of good quality is for that chair, first.
AI outputs information arranged as sentences. A copywriter outputs decisions. Until you understand that distinction, every prompt you write will produce the same flat, passable, forgettable result.
The distinction matters because it changes what you bring to the tool. Most people bring a topic. A good copywriter brings a target, a position, a tone, a single thing that needs to shift, and a clear idea of what the copy needs to do next. The moment you start bringing those things, the output changes.
This guide is about building that habit. Not tips. Not templates. A mental model that changes how you work with AI permanently.
Here is the same brief given 2 ways. The brief: a Lagos-based restaurant is launching a weekend brunch menu.
A copywriter's function is not clever phrases or catchy taglines. It is a series of invisible decisions made before a single word is written.
Most people think copywriting is about language. It is actually about sequence. A copywriter asks a specific set of questions before touching a keyboard, and those questions determine everything.
Not "young women who like skincare." That is a demographic. A copywriter needs the specific person: she has tried 4 moisturisers in 6 months, each one left her skin feeling tight by noon, and she is tired of spending money on promises. That specificity changes every word choice.
Copy that ignores existing beliefs fails. If your audience already believes "natural skincare does not actually work," your copy cannot open with "100% natural." It has to address the belief first. A copywriter knows this. AI, given a blank prompt, does not.
Every piece of copy has a single job. Not five jobs. One. Move the reader from one state to another. Skeptical to curious. Unaware to interested. Interested to ready to buy. Copy that tries to do everything does nothing. The copywriter decides what that one thing is before writing.
Think of copy as a step. A caption leads to a profile. A profile leads to a link. A landing page leads to a purchase. The copywriter always knows what "next" is, because the copy has to earn it. Writing without knowing the next step is writing with no end in mind.
This is the hardest one. Sometimes the most powerful message is the one underneath the words. A luxury brand does not necessarily have to say "this product signals that you have taste." But every word, every image, every spacing decision has to make the reader feel it. Subtext is a copywriter's tool. AI does not introduce subtext without explicit instruction.
Think of a copywriter like a surgeon, not a typist. The work that matters happens before the procedure's first cut — understanding the patient, reading the scans, deciding what to do and what not to touch. The writing is the procedure. The decisions are the diagnosis. This guide is about learning to diagnose.
Two businesses selling the same thing: a productivity app. Same features. Different copywriters.
AI is the fastest first draft you have ever had access to. You are still the writer. These are the inputs only you can provide.
There is a persistent myth that AI will eventually replace copywriters. It will not, for a specific reason: AI can generate, but it cannot judge. It does not know whether the output is right for this person, at this moment, for this brand. That judgment belongs to you.
Here is what AI cannot supply from a blank prompt:
AI defaults to a pleasant, neutral tone that is broadly inoffensive. But brand voice is not neutral. Hardline is not pleasant. Innocent Drinks is not formal. Ryanair is not warm. Voice is a decision. AI will imitate a voice if you describe it clearly. It cannot invent one that is true to your brand without you defining what that is.
AI does not know how much is riding on this piece of copy. If this email is going to 40,000 people and it is your first campaign after a brand relaunch, the stakes change every decision — the level of care, the tone, the risk tolerance in the headline. Only you know what is at stake.
What the copy cannot say directly but must make the reader feel. This requires knowing the culture, the audience's unspoken politics, the things that have been promised and not delivered before, and the emotional register that lands in this specific context. AI writing is typically surface-level. Subtext requires depth of context that you hold.
Demographics are a start. But a 34-year-old Lagos woman who runs a small fashion business, has been let down by 2 agencies in 18 months, and is highly online is very different from a 34-year-old Lagos woman who just became a mother and is stepping back from the business temporarily. Same demographic. Completely different copy. Only you — or a copywriter who has done the research — can hold that distinction.
Every industry has things that are true but cannot be stated directly. A high-end real estate brand cannot say "we cater to people who want to signal wealth." A therapy app cannot say "for people whose parents failed them." But the copy can be built around those realities without naming them. That structuring instinct is human. You have to bring it.
Reframe your relationship with the tool. AI is not a copywriter you hired. It is a very fast typist who will write whatever you brief it to write. The quality of the brief determines the quality of the output. Always.
Before you open any AI tool, write down: who this is for (specifically), what they already believe, what one thing needs to shift, and what the copy needs to do next. That list is your brief. Everything AI produces without it is a guess — and most guesses sound the same.
This is not a prompt template dump. It is a thinking framework in the shape of a prompting guide.
Most AI prompts look like this: "Write a caption for my skincare brand." This is not a prompt. This is a topic. A prompt is a brief. It contains the information a writer would need before they could write anything worth reading.
A good prompt has 5 components. Give AI all 5, and the output becomes worth editing. Give it fewer, and you are editing a guess.
Not demographics. The person. What they have tried, what they believe, what frustration or desire is sitting just below the surface. The more specific this is, the more the copy sounds like it was written for someone real.
Is the goal to get a click? Build trust? Create curiosity? Change a belief? Make the reader feel seen? Pick one. If you list three goals, you will get copy that achieves none of them fully.
Do not say "professional." That means nothing. Say: "direct, no warmth signalling, confident without being arrogant, no exclamation marks." Or: "conversational, like a friend who happens to be an expert, never clinical." Give AI a direction it can follow, not a category it has to interpret.
What is the core claim, observation, or feeling this piece of copy must land? Everything else exists to support this one thing. State it explicitly in the prompt. AI will not find it for you — it will average across everything you give it unless you are clear about what matters most.
This is underused. Telling AI what to avoid is as powerful as telling it what to do. "Do not use the word 'journey.' Do not open with a question. Do not use any phrase that could appear in a corporate newsletter." Negative constraints give AI boundaries, and copy written within boundaries is almost always sharper than copy written in open space.
Time spent on the prompt is never wasted. A 3-minute prompt produces a 2-minute edit. A 15-second prompt produces a 45-minute edit — and the result is usually still mediocre, because you are editing a guess. The prompt is where the thinking happens. AI is where it gets typed.
This is where copy with taste separates from copy that sounds generated. Five editorial moves. Repeatable. Applicable to any AI output.
Even with a strong prompt, AI output requires editing. Not because it is wrong — because it is safe. AI defaults to the centre. It avoids anything that might be wrong by avoiding anything that takes a position. Your job in editing is to move the copy from the centre toward something with a point of view.
These are the 5 moves. Apply them in sequence.
AI almost always begins with a sentence that warms up to the point. Cut it. Start where the piece actually starts. "In today's competitive market, standing out has never been more important." Delete. Whatever came next is the real beginning. This one move makes copy 30% sharper before you have changed a single word of substance.
AI gravitates toward abstract claims because they are broadly true. "High quality." "Designed for results." "Built with care." These are placeholders. Replace each one with a concrete detail. Not "high quality" — "stitched to a 4mm seam so it holds after 200 washes." Specificity is proof. Abstractions are claims. Claims without proof are just noise.
This is the editorial move most people skip, because it requires you to know your audience deeply. It means adding a sentence or phrase that names, without stating, the real emotional stakes. The reader should finish the piece feeling understood — not just informed. This is what makes copy feel personal even when it is written for thousands of people.
AI uses transitional sentences to signal it is moving from one point to another. "Now that we understand X, let's look at Y." These sentences do not exist to help the reader. They exist because the model was trained to be coherent. Delete every sentence that summarises what the previous paragraph just showed, or explains what the next paragraph will cover. Trust the reader to follow.
AI endings are almost always weak. They summarise, or they give a generic call to action, or they end on a question designed to seem engaging. ("What does this mean for your brand?") The last line is prime real estate. It is what the reader carries away. It should advance the argument, not restate it. If you can, end on the truest, sharpest version of the one thing your copy was built to land.
Run this after every AI output before you publish, send, or share anything.
A real brief type. An AI first draft. The editing process shown step by step. Every decision annotated.
The business: a Lagos-based brand strategy consultancy targeting growth-stage founders. They are launching a new service: a 2-hour brand audit session. They need a landing page hero section — the opening copy that decides whether a visitor stays or leaves.
Lagos-based founder, growth-stage business, has spent money on branding or marketing in the last 18 months and is not seeing the return they expected. They sense something is wrong but cannot name it. They are high taste, digitally literate, frustrated with generic agencies, and willing to pay for real thinking — but skeptical, because they have been promised results before.
Move them from "skeptical but curious" to "I need to book this." Not persuade in the conventional sense. Make them feel understood well enough that they trust the consultancy can diagnose what they cannot.
Diagnostic, not prescriptive. Certain, not arrogant. No agency-speak. No exclamation marks. No promises about transformation. Direct enough to feel like the consultancy already knows what is wrong.
Most businesses have a thinking problem, not a marketing problem — and this audit is how you find out which one you have.
A sales pitch. A list of features. Any phrase that begins with "Are you ready to." No words: journey, holistic, game-changing, excited, passionate.
Headline: "Unlock the True Potential of Your Brand with a Professional Audit"
Body: "Are you struggling to see the results you deserve from your marketing efforts? Our brand audit service is designed to identify the gaps in your strategy and give you a clear roadmap to success. In just 2 hours, our expert team will analyse your brand from the ground up and provide you with actionable insights to transform your business. Don't let another day pass without understanding what's holding you back. Book your session today."
This is the output when the brief is not fully provided. It has every problem: generic headline, question opener, agency-speak, vague claims, weak ending. We will now edit it using the 5 moves.
"Are you struggling to see the results you deserve from your marketing efforts?" — Delete. This is a question that assumes the reader's pain in a way that feels presumptuous and generic. The body starts where the real idea is.
"Identify the gaps in your strategy" → too vague. "Clear roadmap to success" → cliché. "Actionable insights" → meaningless. Replace with: what the audit actually does. It examines brief quality, audience specificity, and whether the current positioning can answer one question — why would anyone choose you over the alternative?
The reader is frustrated and has been let down before. The copy needs to acknowledge that without saying it. The subtext: "we know the problem is probably not what you think it is." That needs to be embedded in the copy, not stated directly.
"Don't let another day pass without understanding what's holding you back." — Delete. This is a transition to a call to action, not an idea. It assumes the reader has agreed to feel urgency. They have not. The CTA earns urgency; it does not declare it.
"Book your session today" → No. The last line should do work, not issue a command. It should make the CTA feel like the logical, obvious next step for someone who has understood the argument. Rewrite: "If you have been spending on execution and not seeing the return, the audit is where that conversation starts."
Headline: Most businesses do not have a marketing problem. They have a thinking problem.
Body: The brief is usually wrong. The positioning does not answer the one question every buyer is asking before they commit. The strategy is a document, not a diagnosis. These are not creative failures. They are structural ones — and no amount of ad spend fixes a structural problem.
The Hardline Audit is 2 hours. We examine your brief, your positioning, and your current execution. We name what is not working and why. The recommendation is always grounded in the diagnosis — never the other way around.
If you have been spending on execution and not seeing the return, the audit is where that conversation starts.
Book the Hardline Audit.
Every edit is traceable back to one of the 5 moves. No opener. No abstractions. Subtext embedded ("structural failures" speaks to the reader who has blamed execution while the real problem was upstream). No transitions. Last line earns the CTA rather than issuing it.
This guide gave you a model. Here is what it did not give you, and what that means for what comes next.
This guide covered the thinking that sits underneath prompting. The 5 editorial moves. How to brief AI the way a copywriter briefs themselves. And one worked example showing how the model applies to a real landing page brief.
What it did not cover: long-form copy systems. Email sequence strategy. How brand voice guidelines translate into AI prompts at scale. How to maintain consistency across 30 pieces of content produced in a single month. Those are separate disciplines. This guide was intentionally upstream of all of them — because none of those disciplines work until the foundational model is in place.
The marker of having genuinely absorbed this guide is simple: the next time you open an AI tool, you write the brief before you write the prompt. The quality of your output will tell you immediately whether the model has landed.
AI is not the writer. You are. AI is the fastest first draft you have ever had access to. The brief is yours. The judgment is yours. The editing is yours. The tool just types faster.
If applying this feels harder than it should — because the brand voice is not defined, the audience is not specific enough, or the strategy underneath the copy has not been audited — that is not a prompting problem. It is a foundation problem.
We work with businesses that are at that point. The thinking that makes this guide work — audience clarity, positioning, brand voice, strategic diagnosis — is what we build with clients before any execution starts.