Copy That Converts Readers to Customers

Writing Copy That Converts

The Marketing Team That Stopped Guessing

The team started experimenting with ChatGPT Plus in early 2023, not to replace their writers, but to change how they briefed the work. Instead of handing AI a product spec sheet and asking for copy, they built what their lead called 'desire briefs', short documents describing exactly who the buyer was, what they were afraid of, what they secretly wanted, and what objection was most likely to stop them from clicking 'buy.' They fed those briefs into ChatGPT alongside the product details. The output wasn't publish-ready. But it was directionally right in a way that generic copy never was. It had friction built in, it named the doubt before resolving it. Their writers used these AI drafts as starting points rather than blank pages.

Within eight weeks, their average email click-through rate climbed from 2.1% to 4.7%. Landing page conversion moved to 5.2%. The team didn't change their product lineup, their audience targeting, or their design. They changed how they thought about the brief, and then they used AI to execute against that better brief faster. That's the core tension this lesson is about. AI can write fast. It can write grammatically. It can match a tone. But copy that converts isn't just well-written, it's psychologically precise. The difference between AI copy that underperforms and AI copy that converts is almost always found in the quality of the instructions you give before the first word is generated.

The Principle: Brief Quality Determines Copy Quality

What Pepper Content discovered, and what high-performing marketing teams across industries keep rediscovering, is that AI copy output is a direct reflection of your input clarity. This isn't a technical observation about how language models work. It's a practical truth about communication. If you tell a junior copywriter 'write something for our new running shoe,' you'll get something generic. If you tell them 'write for a 38-year-old woman who's done three 5Ks and wants to do her first 10K but is scared her knees can't handle the training,' you'll get something that lands. AI works exactly the same way. The model isn't the variable. Your brief is.

Why Most AI Copy Falls Flat. A Story About Sameness

Picture a marketing manager at a regional insurance brokerage. She's been asked to write copy for a new small business liability product. She opens ChatGPT, types 'write me a landing page for small business liability insurance,' and gets back something that reads like every other insurance landing page on the internet. 'Protect your business. Comprehensive coverage. Peace of mind.' It's not wrong. But it's not right either. It could have been written for any insurance product by any company in any city. There's nothing in it that would make a restaurant owner in Austin, Texas feel like someone actually understands the specific anxiety of a slip-and-fall lawsuit when you're running a 12-table spot with three staff.

The problem isn't the AI. The problem is that generic input produces generic output, and generic copy doesn't convert because it doesn't create recognition. Recognition is the moment a reader thinks 'that's exactly my situation.' It's the psychological trigger that separates copy that gets scrolled past from copy that gets clicked. When you give AI a vague brief, it draws on the most common patterns in its training data. That means you get the average of all the copy ever written on that topic. Average copy performs at average rates. The insurance manager's landing page had a 1.8% conversion rate. Not because AI is bad at copy. Because she asked for average.

The fix is specific and learnable. It doesn't require any technical skill. It requires you to do the same thinking a great copywriter does before they write a single word: Who is this person? What do they already believe? What are they afraid of? What would make them trust you? What's the one thing that would make them act today rather than tomorrow? Once you have those answers, even rough ones, you can feed them into any AI tool and the output quality jumps dramatically. This is what professional copywriters call 'the brief,' and it's the most valuable thing you can build before you open ChatGPT, Claude, or Copilot.

A Different Industry, The Same Lesson: SaaS Sales at Intercom

Intercom, the customer messaging platform, published internal findings in 2023 showing that their sales development team had started using AI to personalize outbound email sequences. But here's what made their approach work where others failed: they didn't use AI to write more emails. They used AI to write fewer, better ones. Each SDR was responsible for a specific vertical, say, e-commerce brands with 50-200 employees. Before writing any outreach copy, they used Claude to analyze publicly available information about their target accounts: recent product launches, job postings (which reveal company priorities), and customer review themes. Claude synthesized this into a one-paragraph 'situation brief' per account. The SDR then fed that brief into a second prompt asking for a three-email sequence tailored to that specific company's likely pain points.

The result was email copy that referenced real context, not fake personalization like 'I noticed you're in the e-commerce space' but genuine specificity like 'your Trustpilot reviews mention slow response times during peak sale periods, that's exactly the problem our queue management feature was built for.' Reply rates on these sequences ran 3x higher than their previous templated outreach. The lesson isn't 'use AI for sales emails.' It's that AI-assisted research feeding AI-assisted writing produces copy that feels human because it's grounded in actual knowledge of the reader. The tool is Claude or ChatGPT. The skill is knowing what information to gather before you start writing.

Comparing Copy Approaches: What Changes When You Brief Better

Click to enlarge

The HR Director Who Fixed a Hiring Problem With Better Copy

Not all converting copy is trying to sell a product. Consider a director of people operations at a 200-person logistics company in the Midwest. They had a persistent problem: their job postings for warehouse operations supervisors were getting applicants, but the wrong ones. People would show up to interviews expecting a desk job with occasional floor time, and the reality was a physically demanding, shift-heavy role with significant team leadership responsibility. Turnover in the first 90 days was high because the job description wasn't converting the right candidates, it was converting warm bodies. The copy was accurate but not honest. It listed duties without conveying reality.

She used Microsoft Copilot (available through her company's existing Microsoft 365 subscription) to rewrite the job posting with a very specific brief: 'Write this for someone who genuinely thrives in physical, fast-moving environments, takes pride in running a tight operation, and gets satisfaction from developing hourly workers into team leads. Be honest about the hard parts, early starts, accountability pressure, constant problem-solving on the floor. We want candidates to self-select in or out before the interview.' The rewritten posting got 30% fewer total applicants. But interview-to-offer conversion went from 18% to 52%, and 90-day retention improved significantly. The copy converted because it was written for a specific person, and it was honest enough to repel the wrong ones.

Putting the Brief Into Practice

Building a strong copy brief doesn't require a marketing degree or a copywriting course. It requires answering five questions with genuine specificity before you open any AI tool. First: Who exactly is the reader? Not 'small business owners', '42-year-old woman who runs a boutique fitness studio, has been open four years, is profitable but tired, and is considering whether to franchise or sell.' Second: What do they already believe about this topic? If they're skeptical of your product category, name that skepticism. If they've been burned before, acknowledge it. AI copy that pretends objections don't exist gets ignored. Copy that names the objection and addresses it directly earns trust.

Third: What's the one emotion driving their decision? Fear of loss, desire for status, exhaustion, ambition, guilt, pride, most purchasing and action decisions are driven by one dominant emotion. Name it in your brief. Fourth: What's the single action you want them to take? Not 'learn more about us', 'click to book a 15-minute demo call this week.' Vague calls to action produce vague copy. Fifth: What would make them trust you enough to act? A specific number, a customer story, a guarantee, a credential? Include that in the brief. These five questions take 10-15 minutes to answer. They're the difference between copy that performs at the industry average and copy that outperforms it.

Once your brief is built, the AI tool you use matters less than you might think. ChatGPT Plus, Claude Pro, and Microsoft Copilot all produce strong copy output when given a well-constructed brief. Claude tends to write in a slightly more natural, conversational register and handles nuanced tone instructions particularly well, useful for brands that want to sound like a person, not a press release. ChatGPT Plus is strong at structured formats: email sequences, headline variants, bullet-point features. Copilot is the right choice if you're working inside Microsoft Word or PowerPoint and want copy that slots directly into an existing document. Pick the tool that fits your existing workflow. The brief is the skill. The tool is just the keyboard.

Click to enlarge

What the Stories So Far Tell Us

1. AI copy quality is determined by brief quality, not by which tool you use. Pepper Content's click-through rate nearly doubled not because they found a better AI, but because they built better briefs.
2. Generic prompts produce generic output. When you ask AI to write 'landing page copy for insurance,' you get the statistical average of all insurance copy ever written, which is exactly what your competitors are also producing.
3. Specificity about the reader creates recognition. Intercom's SDR team achieved 3x reply rates by grounding copy in real, researched context about each target account's specific situation.
4. Converting copy works across all professional contexts, not just product sales. The HR director's job posting conversion improvement came from applying the same psychological precision that direct-response marketers use.
5. The 'repel and attract' principle improves conversion rates by helping the right reader self-select in while the wrong reader opts out, reducing wasted time on both sides.
6. A five-question brief (reader, beliefs, emotion, action, trust) takes 10-15 minutes and produces reliably stronger AI output than any prompt trick or tool upgrade.
7. Tool choice should match your workflow: Claude Pro for nuanced tone, ChatGPT Plus for structured formats, Microsoft Copilot for in-document drafting inside Microsoft 365.

The Tone Problem: Why the Same Message Lands Differently

In 2022, Oatly, the Swedish oat milk brand, ran a campaign that divided marketers sharply. Their copy was weird, self-deprecating, and occasionally mocked the idea of advertising itself. Lines like 'This copy is terrible but the product is great' appeared on packaging. Meanwhile, their competitors ran polished, benefit-led copy about sustainability and taste. Oatly outsold most of them in premium grocery channels. The lesson wasn't that bad copy wins. It was that copy written in a distinct, authentic tone, one that matched their audience's sensibility exactly, consistently outperformed generic 'good' copy. The tension every marketer faces is real: you know what you want to say, but you're not sure how to say it in a way that actually resonates with the specific people you're trying to reach.

Oatly's marketing team didn't stumble into that voice by accident. They had a documented tone of voice guide, and every piece of copy, from carton text to billboard ads, was filtered through it. When AI tools entered their workflow, the challenge became identical to what every marketing team now faces: how do you get an AI to write in your voice, not its default voice? The default AI voice is helpful, clean, and utterly forgettable. It sounds like a press release written by someone who has read too many press releases. Getting past that default is the skill that separates professionals who get real value from AI copy tools and those who use them to produce content that sounds like everyone else.

The principle Oatly's story reveals is this: tone is not decoration. It is targeting. When your copy sounds like your audience thinks, it creates immediate recognition, a subconscious signal that says 'this brand gets me.' AI tools can absolutely write in specific tones, but only if you tell them precisely what that tone is. Vague instructions produce vague results. 'Write this in a friendly tone' produces something blandly cheerful. 'Write this the way a no-nonsense procurement manager explains a vendor decision to her CFO, direct, evidence-led, no fluff' produces something that actually sounds like a real person in a real professional context.

Teaching AI Your Brand Voice

A mid-sized HR software company called Lattice faced exactly this problem in 2023. Their marketing team had grown fast, and new writers kept producing copy that felt technically correct but tonally inconsistent. Some emails sounded corporate; others sounded startup-casual. Customers noticed. The team started using Claude to draft copy, but the early results had the same inconsistency problem, the AI defaulted to whichever style seemed most 'appropriate' for the content type rather than matching Lattice's actual voice. The fix came from a simple but powerful technique: feeding the AI real examples of copy that had already been approved as 'on-brand,' then asking it to analyze what made those examples work before writing anything new.

This technique is sometimes called 'show, don't tell' prompting. Instead of describing your tone in abstract terms, you paste in three to five pieces of copy that nail it, ask the AI to identify the patterns, sentence length, vocabulary level, use of questions, level of formality, how claims are supported, and then write new copy that follows those same patterns. Lattice's team built a prompt template that included two 'gold standard' email examples at the top of every copy request. Output quality improved immediately. New team members used the same template, and suddenly everyone was producing copy that sounded like it came from the same company. The AI became a tone enforcer as much as a copy generator.

The broader principle here applies whether you're a one-person marketing operation or part of a fifty-person team. AI tools have no memory of your brand by default, every conversation starts fresh unless you build context into your prompt. The professionals who get the best results treat their first prompt less like a request and more like a briefing document. They front-load the context: who the brand is, who the reader is, what the copy needs to do, what it should sound like, and what it should never do. That briefing might take two minutes to write. The copy it produces saves two hours of revision.

A Different Industry, The Same Principle: Real Estate Copy That Closes

Shift industries entirely. A boutique real estate agency in Austin, Texas, twelve agents, no dedicated marketing team, was spending roughly four hours per property producing listing copy, neighborhood descriptions, and social media posts. Each agent wrote their own copy, which meant wildly different quality and tone across the agency's brand. The principal broker started using ChatGPT Plus in early 2023 to standardize and speed up the process. The first month was rocky, generic listing copy, overuse of words like 'stunning' and 'cozy,' descriptions that could apply to any house on any street. The copy was faster to produce but no better than what they'd had before.

The shift happened when the broker stopped asking for 'listing copy' and started asking for copy written for a specific buyer persona. Instead of 'write a listing description for this 3-bed house,' the prompt became: 'Write a listing description for a dual-income couple in their mid-30s, no kids yet, who work in tech and value walkability, original architecture, and having space to work from home. They've looked at 12 houses and are tired of compromises. Here are the property details.' The output changed completely. It led with the home office nook and the walkability score before mentioning square footage. It used language that matched how that buyer actually talked about what they wanted. Inquiries on AI-assisted listings ran 30% higher in the agency's informal tracking over a six-month period.

Comparing Prompt Approaches: Vague vs. Specific

Click to enlarge

The Consultant's Angle: Proposals That Win Business

Independent consultants face a copy challenge that's distinct from brand marketing. Their copy, proposals, capability statements, LinkedIn summaries, follow-up emails, has to do something extremely difficult: communicate expertise without sounding arrogant, and communicate confidence without sounding like every other consultant who uses the same words. 'Results-driven,' 'strategic partner,' 'end-to-end solutions', these phrases appear in virtually every consultant's materials and mean nothing to the clients reading them. Sarah Chen, an organizational change consultant based in Chicago, described her pre-AI proposal process as 'three hours of staring at a blank document followed by an hour of deleting things I'd written.' She billed at $300 an hour and was spending that time on copy.

Sarah started using Claude Pro for first drafts of proposals, but her initial results had the same consultant-cliché problem. The fix came from a counterintuitive move: she started prompting the AI to write the proposal from the client's perspective first. She'd ask Claude to 'write a paragraph explaining what a mid-size healthcare company's COO actually worries about when hiring an external change consultant, the real fears, not the polished objections.' That output became her brief. Then she'd ask for a proposal introduction that directly addressed those fears with specific evidence from her past work. The proposal stopped leading with her credentials and started leading with the client's situation. Her close rate on proposals improved from roughly 35% to over 50% within two quarters. The copy hadn't changed her expertise, it had made her expertise legible to the person who needed to say yes.

Moving From Draft to Done: The Revision Loop

The most common mistake professionals make with AI copy tools is treating the first output as a draft to edit rather than a starting point for a conversation. The AI isn't a vending machine, you don't insert a prompt and collect a finished product. It's closer to working with a very fast, very capable junior copywriter who needs direction. When the first draft isn't right, the answer isn't to heavily rewrite it yourself. The answer is to tell the AI specifically what's wrong and ask for a revised version. 'The second paragraph is too formal, rewrite it the way someone would explain this to a colleague over coffee' takes fifteen seconds to type and produces a meaningfully better result than thirty minutes of manual editing.

Effective revision prompts are diagnostic before they're prescriptive. Rather than just saying 'make it better,' identify the specific problem: the opening is too slow, the benefit isn't clear until paragraph three, the CTA is buried, the tone shifts between paragraphs. AI tools respond well to specific criticism because they can act on it precisely. You can also ask the AI to critique its own output, 'What are the three weakest parts of this copy and why?', and use that self-assessment to guide your next revision request. This back-and-forth loop typically produces polished copy in two to three rounds, which is still dramatically faster than a traditional copywriting workflow involving briefs, drafts, and review cycles.

One more technique worth building into your workflow: ask for variations rather than just revisions. Once you have a solid draft, ask the AI to produce two or three alternative versions that take a different angle, a more provocative opening, a question-led headline, a version that leads with a customer outcome instead of a product feature. You're not going to use all of them, but having three real options to compare is more valuable than trying to imagine alternatives in your head. Many marketers find that the 'wrong' version contains a line or phrase that's actually stronger than anything in the 'right' version. AI-generated variations give you raw material to combine and improve, not finished products to choose between.

Click to enlarge

Key Principles From Part 2

1. Tone is targeting, copy that sounds like your audience thinks creates immediate recognition and trust before the message even lands.
2. Show-don't-tell prompting produces better voice consistency than describing your tone in abstract terms, paste real examples and let the AI analyze the patterns first.
3. Every AI copy conversation starts fresh, front-load your prompt with a full briefing: brand voice, audience, goal, format, and constraints.
4. Audience specificity transforms generic copy into copy that converts, the more precisely you define who you're writing for, the more precisely the AI can address what that person actually cares about.
5. The client-first reversal technique, prompting the AI to write the reader's internal monolog before writing the copy, produces proposals and sales messages that meet people where they are.
6. Treat AI copy output as the start of a conversation, not a finished draft, specific revision instructions produce better results than manual rewrites.
7. Ask for variations, not just revisions, three different angles give you raw material to combine, and the 'wrong' version often contains the strongest line.

Making AI Copy Sound Like You. Not a Robot

In 2023, Mint Mobile's marketing team faced a familiar problem. Their brand voice, sarcastic, self-aware, built on Ryan Reynolds' deadpan humor, was hard to replicate at scale. When they experimented with AI-generated ad copy, the first drafts were technically correct but completely lifeless. The jokes landed flat. The tone was 'friendly' in a generic, forgettable way. Their head of content described it as 'someone doing an impression of our brand after reading one tweet about us.' The copy converted poorly. More damaging, it diluted the brand equity they'd spent years building.

The fix wasn't abandoning AI. It was training it, not technically, but editorially. The team built what they called a 'voice brief': a short document containing actual examples of approved Mint Mobile copy, a list of words they never use, and three sentences describing their tone as if briefing a new copywriter. They pasted this brief at the top of every AI prompt. Output quality jumped immediately. The AI stopped writing like a chatbot and started writing like someone who'd actually read the Mint Mobile style guide.

The principle here is simple but powerful: AI writes to the quality of the brief you give it. Vague input produces generic output. Specific, example-rich input produces copy that sounds like it came from your brand. This isn't a technical skill. It's an editorial one, and most marketing professionals already have it. They just need to apply it to how they talk to AI tools.

The Three Layers Every Converting Copy Needs

Joanna Wiebe, the copywriter behind Copy Hackers, spent years studying what separates copy that converts from copy that just sits there. Her research consistently points to three layers: the message (what you're saying), the mechanism (why it works or how it delivers the benefit), and the match (how closely the copy reflects the exact words and fears your audience uses). Most AI-generated copy nails the first layer. It struggles with the second and third, because those require real customer knowledge that you have and the AI doesn't.

The mechanism is the specific detail that makes a claim believable. 'Lose weight fast' is a message. 'Lose weight by eating in a six-hour window, with no calorie counting' is a message with a mechanism. AI will generate the first version unless you explicitly tell it the mechanism. That means your prompts need to include the specific 'how' behind your product or service, not just what it does, but the distinctive way it does it. That's information you hold.

The match layer is where real customer language lives. If your customers describe their problem as 'drowning in admin,' your copy should use that phrase, not 'operational inefficiency.' AI can replicate this voice perfectly once you feed it the raw material. The source? Customer reviews, support tickets, sales call notes, survey responses. Paste three to five direct customer quotes into your prompt alongside your brief, and watch the output shift from polished-but-generic to specific-and-resonant.

Click to enlarge

A Retailer, A Nonprofit, Same Principle

Patagonia's marketing team doesn't sell outdoor gear, they sell environmental conviction, and the gear is almost incidental. Their email copy routinely outperforms industry benchmarks not because it's clever, but because it's specific. A fundraising email about a specific river in Chile. A product description that names the factory and the worker who made it. When Patagonia's team uses AI drafting tools, they feed in those specifics, the named place, the actual statistic, the real story, and ask AI to shape the narrative around them. The result reads like journalism, not marketing.

The Children's Literacy Initiative, a nonprofit in Philadelphia, applied the same logic to grant proposals and donor emails. Their development team started feeding AI real student stories, anonymized but specific, alongside outcome data. Instead of 'we help children learn to read,' their copy started saying 'last year, 73% of our third-graders in North Philadelphia reached grade-level reading, up from 41% four years ago.' AI helped them find the sharpest version of each sentence. The humans provided the facts that made those sentences matter. Donor response rates climbed 22% in one quarter.

What a Sales Manager in Cincinnati Figured Out

Marcus runs a regional sales team for a commercial HVAC company. Not a glamorous content vertical. His team was spending hours writing follow-up emails after site visits, each one slightly different, none of them particularly compelling. Marcus built a simple system: after each site visit, his reps fill in a five-field notes template (client name, main problem observed, specific detail from the visit, proposed solution, one thing the client said they cared most about). That template gets pasted into ChatGPT with a prompt asking for a follow-up email that references those specifics.

His close rate on proposals went up 18% over six months. The emails didn't feel like templates. They felt like the rep had been paying attention. One client told Marcus it was the most 'personalized proposal process' they'd experienced from any vendor. The AI didn't do anything magical. Marcus just understood that specificity is trust, and he built a process to feed the AI the raw material it needed to sound specific. Any sales team, any industry, can replicate this in an afternoon.

Putting the System to Work

The professionals getting the best results from AI copy tools aren't the ones using the most sophisticated prompts. They're the ones who've built simple, repeatable input systems. A voice brief they update quarterly. A copy fuel document they add to weekly. A notes template their team fills in after every client interaction. These aren't technical systems, they're editorial habits that happen to feed an AI instead of a junior copywriter.

Speed matters too. The right workflow is: use AI to generate fast, use your judgment to select, then edit for authenticity. Don't write from scratch when you can react. Ask AI for ten headline options and cross out eight. Ask for three email structures and pick one. Ask for five ways to phrase the guarantee and choose the one that sounds most like you. Reacting and editing is three times faster than generating from a blank page, and the output is often better because you're making choices rather than staring at a cursor.

One final thing: the best AI-assisted copy still needs a human final pass. Read it out loud. If you stumble on a sentence, the reader will too. If it sounds like something no one would actually say, cut it. AI tools like Claude Pro and ChatGPT Plus are excellent at producing solid first drafts, but the sentence that makes someone stop scrolling and click usually comes from a human who knows exactly what their audience is afraid of, hoping for, or tired of hearing.

1. AI writes to the quality of your brief, vague prompts produce generic copy, specific prompts produce brand-aligned copy.
2. A voice brief (tone description + approved examples + words to avoid) is the single highest-leverage input you can give any AI copy tool.
3. High-converting copy needs three layers: message, mechanism, and match. AI handles structure, you provide the specific details for all three.
4. Customer language, pulled from real reviews, emails, and call notes, transforms AI output from polished-but-generic to specific-and-resonant.
5. The right workflow is generate fast, select with judgment, edit for authenticity, reacting to options is faster and often better than writing from scratch.
6. A 'Copy Fuel' document, updated regularly with quotes, stats, and winning phrases, is the operational habit behind consistently strong AI copy.
7. Every AI-generated draft needs a final human read-aloud pass; the sentence that makes someone click usually comes from a human who knows the audience's real fears and hopes.