# Landing Page Copywriting > Writes and edits conversion copy for marketing landing pages - voice-of-customer mining, outcome-first headlines, value props, CTA labels, specific social proof. Use when the user asks to write, draft, rewrite, or fix landing-page copy, a hero headline, CTA text, or messaging. Not for section layout, visual design, or A/B testing. - Skill: `paldom/landing-page-copywriting` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add paldom/landing-page-copywriting` - Raw SKILL.md: https://api.skillmd.com/api/skills/paldom/landing-page-copywriting/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: Paldom (https://skillmd.com/u/paldom) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/paldom/landing-page-copywriting --- # landing-page-copywriting Write landing-page copy that reads as specific and human, not generic AI slop. ## The failure this fixes Asked to write landing-page copy, a model produces **generic, invented, self-referential** lines — abstract slogans ("Innovation, redefined", "Empowering teams to do their best work"), fabricated-sounding proof, and made-up urgency, all in the even-cadence "grammatical symmetry" that readers now consciously flag as AI-written. It commits to nothing because it did no customer research. The fix: **extract real language and state concrete outcomes; never invent claims, stats, testimonials, or scarcity.** ## When to use / when NOT to use Use for the *words* on the page: headlines, subheads, body, CTA labels, social proof, objection copy, microcopy. Not for: the section skeleton / ordering / form-field decisions (`landing-page-structure`), palette and type (`visual-design-system`), running tests (`landing-page-experimentation`), or plain translation (a native task, not conversion rewriting). ## Workflow 1. **Mine before you write (VoC first).** Do not draft from imagination. Pull verbatim phrases from 3-star reviews, support tickets, sales-call notes, and community threads; bucket them into pain / desire / objection / proof / emotional language; the most-frequent pain becomes the hero, in the customer's words (paraphrasing into marketing-speak strips what converts). If no research exists, say so and gather a lightweight sample first — never fabricate to fill the gap. Full process: `references/copywriting-playbook.md`. 2. **Pick a framework by audience awareness, not taste.** Cold/unaware → AIDA; problem-aware/retargeting → PAS; hidden-gap audiences → BAB; skeptical, high-consideration → PASTOR / 4Ps. One page cannot serve all awareness levels — segment. Table + structures in the playbook reference. 3. **Draft block by block, generating variants.** Produce 3-5 options *per block* (headline, subhead, bullets, proof, CTA), not whole pages. Apply the rules below to each. 4. **Self-QA against the slop filter.** For every line ask: does this sound like every other page in the category? Would a competitor say the same thing? Does this sentence add information or just words? Cut anything that fails. Banned vocabulary and AI tells: `references/copywriting-playbook.md`. 5. **Hand off for human edit.** State plainly that raw output should be human-edited before shipping, and that specific % lift numbers from the research are directional, not promises. ## The rules - **Clarity over cleverness.** The hero answers *what is this, who is it for, why care* in 3-5 seconds. ~80% of visitors read only the headline and first subhead. Read at a 5th-7th-grade level. - **Headlines: outcome-first + specific.** Use `[specific outcome] for [specific audience] without [pain]`, anchored with a number or timeframe ("Cut invoice reconciliation from 3 days to 20 minutes"). Statements beat questions. ~4-11 words. Write the headline *last*, after the offer and proof, so it distills the strongest supported claim. Avoid the generator tic `The [adjective] [category] for [audience]`. - **Value prop: benefit-led and concrete.** Translate every feature via "so that…"; replace adjectives with facts ("Works with Salesforce, HubSpot, and Pipedrive - connect in under 2 minutes", not "seamlessly integrates"). - **CTA discipline.** Describe the *outcome of clicking*, not the mechanism: "Start my free trial", never "Submit" / "Learn more". One primary CTA per page; repeat the *same* CTA down long pages rather than adding competing ones. First-person edges second-person. Match commitment to traffic temperature: cold → soft ("See how it works", "Get the guide"); warm → trial/demo; hot → direct buy. Never "Request a demo" for cold traffic. - **Social proof = specificity.** Generic proof ("Great service!", bare star counts, "Trusted by thousands") now *hurts* - it reads as fake. Use attributed, quantified, outcome-specific proof (full name + role + number: "Booked 23 calls in the first month"). Put at least one proof element in the first screen and place each proof *adjacent to the claim or objection it answers* (ROI proof near pricing, security near the form). - **Message match.** Headline, offer, visual tone, and CTA verb must continue the exact conversation the ad or search query started. Fix this before anything cosmetic - it is the single highest-leverage, most under-executed lever. - **Objection handling is structural.** Put a proactive "You might be wondering…" block mid-page near the CTA plus risk-reversal microcopy under the button ("No credit card required"). Cover the 7 recurring objections (reference). - **B2B vs B2C.** B2B = one person justifying to a committee: direct, credentialed, ROI/risk-framed, named case studies, staged CTAs. B2C = single buyer: conversational, aspirational, reviews/UGC, one decisive CTA. Don't apply B2C urgency tactics to risk-averse B2B buyers. - **Mobile is a separate pass.** Trim AI drafts 15-20%; headline + CTA must fit a 390px screen above the fold. - **No fake anything.** Fabricated countdowns, fake low-stock, invented metrics/testimonials are actively harmful to trust. Urgency must be literally true. ## Output Copy that is customer-sourced, specific, one-primary-CTA, message-matched, and free of the banned AI tells - delivered as labeled blocks (headline / subhead / bullets / proof / CTA), with any research % framed as directional and a note that a human edit pass is expected. ## References - `references/copywriting-playbook.md` - VoC mining process, framework table with structures, awareness stages, the 7 objections, the banned-vocabulary / AI-tell list, and the self-QA filter.