---
title: Building ad-matched landing pages with Claude Code
slug: ai-landing-pages-with-claude-code
summary: The workflow Will showed on Dara Denney's channel (Oct 2026) — pick a winning ad angle in Meta Ad Library, feed Claude Code the brand's PDP text, images, fonts, colours and a landing-page playbook, get an ~85–90%-done HTML page in ~45 min, edit by screenshot, deploy to Vercel.
domain: marketing
tags: [landing-page, ai, claude-code, workflow, vercel, cro]
related: [ad-landing-page-message-match, persona-landing-pages-and-listicles, voice-of-customer-copy-mining, writing-advertorials-with-ai]
sources: [yt:2HDkD7jSbzE]
updated: 2026-10-24
---

# Building ad-matched landing pages with Claude Code

## TL;DR
- Landing pages used to take 3–4 weeks (6 for a first page at Thesis, the CRO agency where Dara worked); with Claude Code, a first draft in ~45 minutes, polished in 2–3 hours[^1][^2].
- Start with a **hypothesis**: one ad angle that is already winning — don't run the workflow with your brain off[^3].
- Feed it a "content center": winning ads, scraped product images, full PDP text (accordions expanded), blog/science content, UGC, fonts and colours from DevTools, and a **landing-page playbook markdown**[^4][^5][^6][^7].
- Output: a full HTML page "85–90% there"; edit by pasting a screenshot + one-line instruction; host on Vercel (free tier) on a subdomain[^8][^9].
- Biggest ROI of Claude Code for marketers, per Will: statics from learnings, landing pages and proposals — beyond that is "AI psychosis" territory[^10].

## Details
Will (guest; his playbook is distributed via Dara's site — see caveats) walked Dara Denney ([[people/dara-denney]]), a first-time Claude Code user, through a Seed probiotics page[^3].

**Step by step**
1. *Find the angle.* Meta Ad Library → brand → sort by impressions → filter media type "meme" to see static ads with copy on them (usually bottom-of-funnel converters). Find several ads saying the same thing — for Seed, "poop every day" / "I don't bloat after every meal"[^3][^11].
2. *Collect assets.* Save those ads; bulk-download all site images with an image-scraper Chrome extension (216 images here); optionally connect an image-generation model for gaps[^12][^4].
3. *Copy the PDP.* Expand every accordion (benefits, ingredients), Cmd+A, paste into Claude Code unprompted. Do the same for educational/blog pages[^4][^5].
4. *Brand styling.* Inspect the homepage headline in DevTools, filter computed styles by "font", then by "color", paste both — this gives text hierarchy and palette[^6][^7].
5. *Run the playbook.* Drop the playbook .md in the project folder, paste its path, say "run this playbook". Will built it by having Claude learn from hundreds–thousands of landing pages; it has checkpoints (page type: conversion / listicle / quiz; copy direction)[^7][^13][^14].
6. *Choose direction.* Offered "bold ad-verbatim" vs "relief + science" — they chose ad-verbatim; they often take the top ad's copy verbatim as the headline ([[ad-landing-page-message-match]])[^15].
7. *Review & edit.* Page self-scored 93/100; edits = screenshot of the area + instruction ("make this headline one line, full block")[^16][^17].
8. *Make it buyable.* Page includes reviews, "risk-free for 30 days", FAQ and a native buy section; no exits — every CTA scrolls back to the buy box. Wire real Shopify product IDs (or the Shopify MCP) for checkout[^18][^19].
9. *Deploy.* Vercel, noindex for previews; production on a subdomain like `poop.brand.com`[^9][^16].

## Disagreements / caveats
- Health/wellness content triggered model safety flags mid-run; they switched model and continued[^13]. Make sure claims stay compliant.
- Copy is where AI is weakest — review it against VOC ([[voice-of-customer-copy-mining]])[^20].
- A run used ~50k tokens and ~30 minutes. The playbook is a gated download (lead magnet), and "Will" is identified only by first name in the video.
- Model names in the video (Opus/"Fable") reflect Oct 2026 Claude Code; the workflow is model-agnostic.

## See also
- [[persona-landing-pages-and-listicles]] — Dose's no-code builder alternative

[^1]: yt:2HDkD7jSbzE @ 00:49 — Will (guest) and Dara Denney
[^2]: yt:2HDkD7jSbzE @ 17:15 — Will (guest on Dara Denney)
[^3]: yt:2HDkD7jSbzE @ 03:18 — Will (guest on Dara Denney)
[^4]: yt:2HDkD7jSbzE @ 05:34 — Will (guest on Dara Denney)
[^5]: yt:2HDkD7jSbzE @ 06:19 — Will (guest on Dara Denney)
[^6]: yt:2HDkD7jSbzE @ 07:56 — Will (guest on Dara Denney)
[^7]: yt:2HDkD7jSbzE @ 08:41 — Will (guest on Dara Denney)
[^8]: yt:2HDkD7jSbzE @ 11:47 — Will (guest on Dara Denney)
[^9]: yt:2HDkD7jSbzE @ 11:01 — Will (guest on Dara Denney)
[^10]: yt:2HDkD7jSbzE @ 15:36 — Will (guest on Dara Denney)
[^11]: yt:2HDkD7jSbzE @ 04:04 — Will (guest on Dara Denney)
[^12]: yt:2HDkD7jSbzE @ 04:49 — Will (guest on Dara Denney)
[^13]: yt:2HDkD7jSbzE @ 10:16 — Will (guest on Dara Denney)
[^14]: yt:2HDkD7jSbzE @ 12:32 — Will (guest on Dara Denney)
[^15]: yt:2HDkD7jSbzE @ 14:02 — Will (guest) and Dara Denney
[^16]: yt:2HDkD7jSbzE @ 16:24 — Will (guest) and Dara Denney
[^17]: yt:2HDkD7jSbzE @ 18:01 — Will (guest on Dara Denney)
[^18]: yt:2HDkD7jSbzE @ 18:47 — Will (guest on Dara Denney)
[^19]: yt:2HDkD7jSbzE @ 19:34 — Will (guest on Dara Denney)
[^20]: yt:2HDkD7jSbzE @ 21:52 — Will (guest on Dara Denney)