You launch Tuesday. The page does not exist yet. This tutorial takes an indie founder, SaaS product owner, or course creator from “I have a thing to sell” to a live landing page — hero, benefits, social proof, FAQ, CTA — in about 90 minutes. AI writes the first draft of every text block, a no-code builder handles the layout, and a real A/B test settles which version wins. The target is a defensible v1 that converts, not a perfect page.
TL;DR
- Time budget: ~90 minutes. Brief 10 min, hero variants 15, supporting sections 20, assembly 30, visual sourcing 15.
- Draft with: ChatGPT (GPT-5.5, free tier works), Claude (Sonnet 4.6, free tier works), or Gemini 3.1 Pro. All three draft a full page in one prompt as of June 2026.
- Ship on: Carrd ($19/year Pro Plus) for a single page, Framer (free to start, $10/mo Basic for a custom domain) for anything that needs to grow.
- Win condition: A/B test one metric for 7–14 days or ~1,000 visitors per variant, decide at 95% confidence.
- Never let AI fake: testimonials, customer logos, or photos of “your team.” Fabricated trust signals tank conversion.
The median SaaS landing page converts at about 3.8%, while custom pages hit 11.6% versus 3.8% for templates (daydream, 2026). AI gets you to a custom-feeling page fast; the test tells you whether it beats what you had.
What this covers
The full draft-to-A/B-test loop: writing the audience brief, generating hero variants, drafting the supporting sections, assembling in a no-code builder, and running a 7–14 day A/B test. AI’s job is the first draft of every text block. Your job is to keep the message tight and the trust signals real.
This is for indie founders launching v1, SaaS teams iterating an underperforming page, course creators with a single product page, and freelancers shipping client pages weekly. Skip it if you already have a copywriter who nails the voice — AI is a faster draft, not a better one. Strong signals you should use it: conversion below your category benchmark (under ~3.8% for SaaS), bounce above 70%, or copy untouched for six months. Weak signal: traffic too low to measure. Get traffic first; AI cannot fix a discovery problem.
Pick your drafting model
All three frontier chat models draft a full page from one brief. The differences that matter for copy work, as of June 2026:
| Model | Free tier | Paid entry | Notes for landing-page copy |
|---|---|---|---|
| ChatGPT (GPT-5.5) | Yes, tight limits | Plus $20/mo | Strongest at punchy, varied hero lines; US free tier shows ads |
| Claude (Sonnet 4.6) | Yes, limited | Pro $20/mo | Cleanest long-form sections; least prone to hype adjectives |
| Gemini 3.1 Pro | Yes (Google AI) | AI Pro $19.99/mo | 1M-token context; good when you paste a long brief plus competitor pages |
For this workflow the free tier of any one of them is enough — you are generating a few hundred words, not a book. If you already pay for one, use it. Do not pay for a second model just for a landing page.
Pick your builder
| Builder | Free plan | Entry paid | Best for |
|---|---|---|---|
| Carrd | Yes (carrd.co subdomain) | Pro Plus $19/year | Single-page sites, fast launch campaigns |
| Framer | Yes (no custom domain) | Basic $10/mo annual | Early-stage pages that will grow; built-in AI sections |
| Webflow | Yes (1 page, .io subdomain) | Site plan $14/mo; CMS $23/mo | Growth-stage marketing sites that scale |
Pricing as of June 2026, verified against vendor and review pages (Carrd/Webflow/Framer comparison; Framer pricing). Carrd Pro Plus at $19/year is the cheapest credible option for a one-page launch. Framer’s free plan blocks custom domains, so budget the $10/mo Basic the moment you go live. Webflow only earns its price once you need a CMS-backed marketing site.
Before you start
- Identify the actual target user — not “founders,” but “founders of 5–10 person B2B SaaS who spend too much on copywriters.” Specificity is the one thing AI cannot invent for you.
- Write the top user outcome in one sentence: “Ship landing page copy in a day, not a week.” This becomes the hero promise.
- Name your one real differentiator versus the top three alternatives. Without one, no headline works. AI papers over this with adjectives — do not let it.
- Collect three real testimonials, three real screenshots, and your real pricing. Fabricated trust signals destroy conversion.
Step by step
- Write the brief (10 min). Target user, top outcome, one differentiator, the top three objections you actually hear. Paste it as one block into your first prompt — these models perform better with all context up front.
- Generate hero variants (15 min). Ask for 10 headlines plus 5 subheads, and explicitly request variety: outcome-led, fear-led, mechanism-led, identity-led. Score each on specificity, length under 12 words, and whether it carries the differentiator. Pick the top two to A/B test.
- Draft supporting sections (20 min). In one request: 3 benefit bullets (outcome-led, not feature-led), social-proof framing (what to wrap around the real testimonials), a 5–6 question objection-based FAQ, and CTA copy (button label plus microcopy underneath).
- Write a visual brief per section. Ask the model what image goes where, what it shows, and why. Source the images separately. Do not use AI-generated photos of people; they wreck the trust signal even when convincing.
- Assemble in your builder (30 min). Carrd, Framer, or Webflow. Use real customer logos and real screenshots. The less “designed” the page looks while staying clean, the better v1 converts — custom-feeling pages convert about 3x better than obvious templates.
- A/B test. New hero versus the current page. Pick one metric — signups, demo bookings, or purchases. Decide in 7–14 days or ~1,000 visitors per variant at 95% confidence, whichever comes first. If neither variant wins decisively, rewrite; do not split the difference.
The hero prompt that works
Reuse this verbatim, swapping the brief:
You are a direct-response copywriter. Here is my brief:
- Target user: [who, specifically]
- Top outcome: [one sentence]
- One differentiator vs [top 3 alternatives]: [the difference]
- Top 3 objections I hear: [list]
Write 10 hero headlines (each under 12 words) and 5 subheads.
Give me variety: outcome-led, fear-led, mechanism-led, identity-led.
For each headline, note its angle. No hype adjectives, no "revolutionary."If the third pass finally lands a headline, that is the prompt you keep. Save it.
Quality check
- Read the page top to bottom as your target user. Does the hero promise survive contact with the benefits and FAQ? If sections contradict, you have two messages — pick one.
- Open it on mobile. The hero is the first thing ~70% of visitors see; if it is buried under a hero image, the headline does not matter.
- Read it aloud. AI copy smooths into corporate-speak when you skim. Reading aloud forces you to cut the filler adjectives.
- Run a five-second test on a friend. They should repeat the value prop after five seconds. If not, the hero is broken.
Common mistakes
- Launching “perfect” on day 1. Ship v1 in 90 minutes; iterate on data.
- AI-generated team or customer photos. A trust-signal killer — visitors detect them instantly.
- No A/B test plan. Without a test you have an opinion. With one you have a decision.
- A hero promising what the product cannot deliver. Bounce spikes when the rest of the page contradicts the headline.
- Five different CTAs. Pick one primary action and repeat it three times down the page.
- Skipping the FAQ. The FAQ is where you handle objections; without it, conversion bleeds at the bottom of the page.
- Testing on too little traffic. Under ~10,000 visitors/month, only a >30% lift reads as real. If you cannot hit ~1,000 visitors per variant, ship the stronger draft and revisit when traffic grows.
FAQ
- Which model should I draft in?: Any of GPT-5.5 (ChatGPT), Sonnet 4.6 (Claude), or Gemini 3.1 Pro. Free tiers handle a landing page. Sonnet 4.6 tends to write the cleanest sections; GPT-5.5 the punchiest heroes. Do not pay for a second model just for this.
- Should I write in AI or in my builder?: Draft in AI as one block, then paste into the builder section by section. Mixing the two mid-draft loses paragraphs.
- How long should the hero headline be?: Under 12 words, ideally 5–8. Every model defaults to longer; cut.
- Can AI write the testimonials?: No. Use real customer words verbatim, even if awkward. Fix grammar, never invent. The median SaaS page already converts at only ~3.8%; fake proof drops it further.
- What does v1 cost?: $0–$20. Free chat-model tier plus Carrd Pro Plus at $19/year, or Framer free to draft and $10/mo Basic once you attach a domain.
- How long should the A/B test run?: 7–14 days or ~1,000 visitors per variant at 95% confidence, whichever comes first. One full week minimum to cover the weekly cycle. Shorter is noise; longer costs opportunity.
- What if my product is hard to explain in a hero?: Lead with the outcome, not the mechanism. “Cut copywriting time by 80%” beats “AI-powered copy assistant.”