<aside> 🚀

Start here

The live example: Limora AI - this is the exact structure we shipped on limora.ai in one day.

The build tool: Claude Code - it reads your existing landing page and writes every new page from the copy that is already there.

Part 1 - The site map: the four page types, and why each one exists

Part 2 - Audience pages, and the first-two-lines rule that gets you quoted

Part 3 - Workflow pages, one URL per question people actually type

Part 4 - The Learn page, and the one rule that decides whether it works

Part 5 - llms.txt, the file that tells AI which pages matter

Part 6 - Backlinks, because structure alone is not enough

Part 7 - The full checklist, and the prompts to build it

</aside>

<aside> ✋

I am not an SEO expert. This is our plan, not received wisdom. We had a landing page that converted and nothing else, so people arriving from Google or from ChatGPT had no page to land on that answered their question. This is what we built to fix that, written down so you can copy the structure. Treat it as a starting point and measure your own results.

</aside>


What SEO and GEO mean here

SEO is what you already know. You want Google to rank your pages for the things people search.

GEO stands for generative engine optimization. You want ChatGPT, Claude, Perplexity and Google's AI answers to mention your product when someone asks them for a tool that does what you do. That recommendation is a referral from a source the person already trusts, which is why it converts so well.

The useful part is that both goals want the same thing from you. They want clear pages, one topic per URL, and answers written in plain language near the top. You are not building two sites. You are building one site properly.

<aside> 🎯

Why this matters more than it used to. AI can build your product now, so the product is no longer the hard part. Distribution is. Every hour you spend on the site is spent on the part that is actually scarce.

</aside>


Part 1 - The site map

We went from a single landing page to a full site in one day. That is possible because a well-written landing page already contains most of the content you need. You are not writing a site from nothing. You are splitting and expanding what you have.

Four page types, in the order we built them:

Page type One per What it is for
Audience page Customer type Someone arrives knowing who they are, not what your feature is called. This page says what you do for them specifically.
Workflow page Job the product does Each workflow is a phrase somebody types into Google or ChatGPT. Each one gets its own URL so it can rank.
Learn page Site The hub for guides and articles. This is what gets referenced over time, and what earns links.
llms.txt Site A plain file at the root that tells AI crawlers which pages matter most.

Here is the structure we used, so you can map your own onto it:

/                              landing page (unchanged)
/pricing

/for/ecommerce                 audience page + video + FAQ
/for/designers
/for/marketers
/for/developers

/use-cases/product-photos      workflow page + FAQ
/use-cases/hero-images
/use-cases/icons
/use-cases/illustrations
/use-cases/mockups
/use-cases/logo-variations
/use-cases/image-animation

/learn                         guides hub
/learn/[article-slug]

/llms.txt

<aside> 💡

Start with four audience pages and three workflow pages. Seven pages is enough to learn what works. Building twenty thin pages before you know which ones get traffic is how this turns into a month of work with nothing to show.

</aside>


Part 2 - Audience pages

▸ The rule that matters: the first two lines