Copywriting Framework: Conversion Copy for Every Page

Created by Corey Haines (compiled from older copywriting rules) - https://github.com/coreyhaines31/marketingskills

Overview

The copywriting framework is a set of working rules for conversion copywriting: website and email copy whose job is to get a specific reader to take a specific next step. Wikipedia defines copywriting as writing persuasive text for advertising or other forms of marketing. This framework narrows that to the pages a software or service company actually ships (homepages, landing pages, pricing pages, feature pages and email sequences) and gives each one a short list of rules to write and review against. It applies to B2B copywriting and sales copywriting as much as to consumer sites, and it suits copywriting for beginners because every rule can be checked against the draft.

The version on this page follows the copywriting skill that Corey Haines published in his open-source marketingskills repository, a free collection of marketing instructions for AI coding agents. The repository's README describes Conversion Factory as Haines's agency for conversion optimization, landing pages and growth strategy. His copywriting skill sets out five principles (Clarity Over Cleverness, Benefits Over Features, Specificity Over Vagueness, Customer Language Over Company Language, and One Idea Per Section), a formula for call-to-action buttons, and separate guidance for each page type. Hamster's page arranges those rules into an order of work and adds the research behind each one.

None of the individual rules is new, and the framework does not claim to be. Written advertising formulas are more than a century old. The attention, interest, desire and action sequence is commonly attributed to E. St. Elmo Lewis, and C. P. Russell first used the AIDA acronym in 1921, according to Wikipedia's history of AIDA. Claude Hopkins argued that advertising should be tested and measured in Scientific Advertising, published in 1923. The framework's contribution is a practical sequence and a checklist that a writer, an editor or an AI agent can apply one page at a time.

Research on how people read web pages explains why clarity carries so much weight. In Jakob Nielsen's study of how users read on the web, 79 percent of test users always scanned a new page and only 16 percent read word by word. A related NN/g experiment found that a version of a site rewritten to be concise, scannable and objective scored 124% higher in measured usability than a promotional version full of exaggeration and boasting. Readers who scan decide from headlines, first words and button labels, so those are where the framework puts most of its effort.

The framework works in two layers. The first layer is a set of principles that apply to every line: be clear, be specific, lead with what the reader gains, use the reader's own words, back claims with honest proof, and label every button with what happens after the click. The second layer is page-specific. Haines's guidance gives a homepage the job of serving several audiences with clear paths for each, gives a landing page a single message and a single call to action, and gives a pricing page the job of helping visitors choose a plan. Email copywriting carries the same rules into the inbox, where his emails skill gives each email one job and one main call to action.

Measurement closes the loop. A rewrite is a hypothesis until real visitors respond to it, so the framework ends with testing headlines and calls to action and feeding what you learn back into the next draft. Benchmarks give rough context: Unbounce's 2024 Conversion Benchmark Report puts the median landing page conversion rate across industries at 6.6% and reports that pages written at a 5th to 7th grade reading level converted at 11.1%. Judge your own rewrite against the page's previous version, because traffic source, offer and price move conversion rates more than any benchmark can capture.

Core Principles

Clarity Over Cleverness

A first-time visitor should understand what you offer, who it is for and what to do next without decoding anything. Haines's copywriting skill states the rule as a choice: if you have to pick between clear and creative, pick clear. NN/g's guidance on how users read on the web makes the same call for sub-headings, asking for meaningful ones and warning against clever ones. Wordplay can stay when it survives a comprehension check. A quick way to run that check is a five-second test, in which people see a page briefly and then say what it was about and who it was for.

Specificity Over Vagueness

Replace general claims with the task, the timeframe, the named outcome or the concrete scenario a reader can picture and check. Haines contrasts a vague line about saving time with one that names the report being cut and the time it used to take. Specific claims have to be true, and the NN/g study that tested writing styles found objective language, free of exaggeration and boasting, improved measured usability on its own (Morkes and Nielsen). If a number would help but you do not have one, describe the scenario instead of inventing a figure.

Benefits Over Features

A feature says what the product does, and a benefit says what the customer gets from it. The best-known version of the idea is the saying that people want a quarter-inch hole, and Quote Investigator traces its earliest known appearance to a 1942 insurance advertisement run by C. C. Wagner. Leo McGivena may be credited with popularizing it, and Theodore Levitt, to whom it is often credited, himself attributed it to McGivena. Benefits persuade better when the feature stays beside them as the reason to believe. Haines's guidance for feature pages follows the same chain: connect the feature to the benefit and the benefit to the outcome.

Customer Language Over Company Language

The words customers use in reviews, interviews and support tickets are usually clearer than the words the product team uses internally. Haines's customer-research skill asks for exact quotes rather than paraphrases. A paraphrase tends to turn a vivid customer phrase back into company language. Joanna Wiebe of Copyhackers teaches Amazon review mining as a reliable way to find messages that land with prospects. The research is the raw material for headlines, subheads and button labels.

One Page, One Job

Each page type serves a different visitor at a different point in the decision. Haines's page-specific guidance gives the homepage the job of offering clear paths for different visitor intents, the landing page one message and one call to action matched to its traffic source, and the pricing page the job of answering which plan is right for the visitor. NN/g's homepage guidelines add that a homepage should open with a one-sentence tagline that says what the company does. Writing every page to the same template produces copy that fits none of them well.

Calls to Action That Say What Happens Next

A button label is a small promise about the next screen. Haines lists Submit, Sign Up, Learn More, Click Here and Get Started as weak labels and offers the formula of an action verb plus what the visitor gets, with a qualifier if needed. NN/g's research on Get Started buttons found the phrase ambiguous enough to send people into flows they did not expect, and its article on Learn More links recommends descriptive, front-loaded labels. A clear label also filters out visitors who would click and then leave.

Honest Proof

Proof makes specific claims believable, and false proof destroys the trust the rest of the page builds. Haines's skill lists honesty among its style rules and warns that invented numbers or quotes damage trust and can expose the company legally. The FTC's Endorsement Guides require that endorsements reflect the endorser's honest opinion and that ads featuring exceptional results disclose what people can generally expect. Use real customer quotes with permission, real counts you can defend, and plain statements where you have no proof yet.

Older Formulas Behind the Framework

The framework borrows from older copywriting formulas. The table shows where each one appears in the steps below, so a writer who already knows one of them can map it across.

Treat these formulas as checklists for the argument a page makes. They do not predict behavior: the Wikipedia article on AIDA summarizes research finding the model a poor predictor of actual consumer behaviour, with little empirical support for hierarchical models of this kind. The Haines templates present Problem-Agitate-Solve without naming an originator, and this page does not assign one.

Steps

1. Collect customer language Before drafting, gather how customers describe the problem, the outcome they want, the alternatives they tried and the doubts that slowed them down. Good sources are sales call notes, interviews, support tickets, survey answers, and reviews of your product and competing products. Copyhackers' review mining method uses targeted searches for phrases such as "tired of" in reviews across your category. Record exact quotes with their source, grouped by theme. This file feeds every later step.

2. Define the page's job and the reader's starting point Write one sentence that says what the page must get the reader to do. Then note where the traffic comes from and what readers already know when they arrive, two of the context questions in Haines's copywriting skill. A visitor from a comparison search needs different copy from one who clicked an ad about a specific problem. If you cannot write the job in one sentence, the page is probably trying to do two jobs and should be split.

3. Write the headline and subhead Draft many headline options from the customer language, then cut to the ones that say what the reader gets in words the reader would use. Check each against NN/g's finding that people scan the first words of headlines and links, and move the information-carrying words to the front. Pair the chosen headline with a subhead that adds who it is for, how it works or what makes it different. Keep two or three alternatives with a one-line rationale each, as Haines's output format suggests, so they can be tested later.

4. Translate features into benefits List the features relevant to this page. For each one, ask what it lets the customer do, and then what that means for their work or life, until you reach an outcome the customer would name. Keep the feature in the copy as evidence for the benefit. Check each benefit against the customer language file, and drop benefits no customer has mentioned or would recognize.

5. Draft the body in the page's section order Lay out the sections the page type needs. For a landing page, Haines's list of core sections covers social proof, the problem, the solution and its benefits, how it works, objection handling and a final call to action. Give each section one idea, which also matches NN/g's advice of one idea per paragraph for readers who scan. Put the most important point first in each section, since many readers never reach the rest.

6. Edit for clarity and add proof Cut the draft hard, replace jargon with the customer's words, and prefer active verbs, as Digital.gov's plain language guide recommends. Read it aloud and rewrite any sentence you stumble over. Then go through every claim and attach proof where you have it: a customer quote, a count you can defend, a named integration, a guarantee. Where you have no proof, make the claim smaller rather than inventing support.

7. Write the calls to action Label every button with the action and what the visitor gets, and add a line of microcopy when the visitor needs to know what happens next, such as whether a card is required. Follow the GOV.UK guidance on links, which asks for descriptive, front-loaded text over generic labels like "click here". Make the primary action on each page obvious and keep secondary actions quieter. Check that the button promise matches what the next screen actually shows.

8. Test and iterate Publish, record the current conversion rate for the page's one job, and test the element most likely to move that rate first, usually the headline or the main call to action. Change one element per test so the result points at a cause. Decide the sample size before the test starts, because Evan Miller shows that stopping as soon as a result looks significant makes the reported significance meaningless. Add what each test teaches you to the customer language file.

When to Use

• You are launching or rewriting a website and need homepage, landing page, pricing and feature page copy that each do a defined job, with one shared set of rules for everyone who writes.
• Traffic is healthy but sign-ups, demo requests or purchases are low, and visitors' comments or five-second test answers suggest they do not understand the offer.
• Several people or AI agents write copy for the same product and you need a checklist that makes their drafts consistent and reviewable.
• A founder or engineering team wrote the current site in internal product terms, and you want to rewrite it in the customer's words.
• You run copy tests and want hypotheses that come from a principle and customer research instead of guesses.

When Not to Use

• The piece is long-form thought leadership or brand storytelling, where voice, argument and narrative matter more than a single conversion action.
• The product has not yet found buyers who want it. Clearer copy will make a weak offer easier to understand but will not make it wanted, so fix the offer first.
• The copy is in a regulated field such as health, finance or law, where legal review of every claim takes precedence over conversion rules and specific benefit claims may not be allowed.
• The readers are domain experts reading documentation or technical references. NN/g's research on writing for domain experts found they want facts, citations and precise shared vocabulary, which benefit-led marketing copy tends to strip out.