The Shelf / Growth / Product SEO Site Builder
Product SEO Site Builder
A website for your product with one page per thing people search — and only for searches you can win.
The job: build your product the website Google can actually rank — a home page, one guide page for each thing people search, and an FAQ wired into those guides. Someone types the problem. They find the thing that fixes it.
Most products have a store listing and maybe a one-page site. Neither shows up in a Google search. Meanwhile the questions your buyers actually type — "how to
What it makes your agent do
- Check Google first, write second. Before a single page is written, the agent searches the exact phrase and reads page 1. All big publishers and official docs? Dropped, and written down so nobody researches it again. Forum threads, or nothing on the mark? That phrase earns a guide. Fewer real pages beat a pile of padded ones, and the skill enforces that trade.
- Write pages built to survive Google's helpful-content system. Same shape every time: the heading is the question, the answer lands in the first 150 words, then a real step-by-step that helps even someone who will never buy from you. The pitch for your product is capped at roughly a fifth of the page, with one real screenshot.
- Wire up the plumbing that helps pages rank. A different title and description on every page. One web address marked as the real one for each page. The preview card that shows when someone shares a link. The hidden fact blocks Google reads — your product's details, the questions and answers, the article, the how-to steps, the breadcrumb trail — all templated in the package. And links running home into the guides, guide into guide.
- Launch without breaking anything. Point old addresses at the new ones, settle on one web address, hook up Google's own reporting tool, hand over the list of pages, and ping the search engines. Plus the warning that saves you a panic: this takes weeks, and the "couldn't fetch" message Google shows right after you hand over the page list is normal.
- Refuse the shortcuts that get sites punished. No made-up numbers or review counts anywhere. No fake star-rating markup. No two pages chasing the same search. No two web addresses serving the same site.
Who it's for
You built an app or a tool and Google shows nothing for it. You'd rather earn traffic that keeps coming than rent installs. Or you build these sites for clients and want the same method every time.
Why it beats a free directory download
Free SEO prompts write pages. They don't know which pages deserve to exist. The Google check, the page shape, the fact-block templates and the launch checklist here come out of shipping real guide sites for real products — with the mistakes already paid for: pages bouncing between two addresses forever, security certificates not ready in time, and the pull to write pages nobody should read. Kept current as Google moves. The field report shows you a real search plan and its verdict before you pay.
FIELD REPORT real output, not a promise
A worked slice from the skill's own documented example — a voice-logging feature of a budgeting app going through Step 2 (keyword plan + SERP gate) and the guide anatomy that follows. No rankings claimed; indexing verdicts belong to Search Console weeks later.
Keyword plan entry (the Step-2 deliverable format, filled)
| # | Keyword (exact) | Intent | Guide must answer | SERP winnability |
|---|---|---|---|---|
| 1 | log expenses by voice | wants hands-free logging | how voice logging works, platform options, assistant vs app | page 1 has 2 forum threads + thin listicle — OPEN |
The verdict column is the gate: two forum threads and a thin listicle on page 1 means a genuinely helpful page from a real site can compete. Had page 1 been official support docs and major-publisher roundups, the keyword would land in the ## Dropped log with the reason — and never get a page.
What that keyword's guide must look like (anatomy enforced by the skill)
- H1, humanized: "How to Log Expenses by Voice (Hands-Free, in Seconds)" — the query, not a clever headline.
- Direct answer in the first 150 words. Zero throat-clearing; this is the text AI Overviews and featured snippets pull. "In today's fast-paced world" is banned on sight.
- Generic step-by-step solution that honestly serves someone who will never install the app — the section that earns Google's trust.
- The product bridge, capped: one H2 ("The one-tap way"), 2-4 sentences in the app's voice, ONE real screenshot of the app doing exactly this. Never more than ~20% of the page.
- Mini-FAQ, related guides with keyword anchor text, CTA band.
800-1500 words, written for a human first.
The schema that ships with it
Article JSON-LD on the guide; FAQPage markup only for questions visible verbatim on the page; aggregateRating only if pulled from a real ratings source with a meaningful count — the skill treats fabricated ratings markup as a manual-action magnet, not a growth hack.
The honest expectation set at handoff
Submit the sitemap, then leave it alone: days-to-weeks to index, check Search Console at ~2 weeks. The deliverable is a site where every page targets a query it can plausibly win — the rankings themselves are Google's call, and the skill never claims otherwise.
SERVICE RECORD living gear — updated as the factory learns
v1.1.0 — 2026-08-25
Field update from a month of shipping and diagnosing real guide sites. New: the "discovered ≠ crawled" guardrail — how to spot guide pages Google records but never fetches ("Discovered – currently not indexed", Last crawled: N/A), why more content can't fix it, and why in-body internal links plus Request Indexing can. The two-edit rule for any copy that also lives in a page's JSON-LD, with the parse check that catches a silently broken block. An /llms.txt step for AI answer engines. Three Search Console gotchas: the property-format error that masquerades as missing permission, the full-URL requirement when resubmitting a sitemap, and the sitemap "Last read" date that exposes a stalled fetch.
v1.0.0 — 2026-07-17
First issue. Ported from the factory's internal skill: sanitized for general use, methodology intact, field report captured from a real run.
Every update ships free to owners — your locker always serves the latest version.
QUESTIONS
Is this just 'generate 50 AI blog posts'?›
No — the opposite. Every phrase has to survive a live Google check before it earns a page. Every page has to genuinely help someone who will never buy from you. The skill ships fewer pages on purpose. Thin mass-produced pages are exactly what Google buries.
Does it work for products other than iOS apps?›
Yes. It was worked out on app sites, but nothing in it is app-only — apps, software, tools, services all run the same way. Pulling ideas from a store listing is just one way to find search phrases. The Google check and the page layout don't change.
What does it build the site with?›
Any plain, fast website setup. It suggests Astro, which is a proven fit, and the launch steps work wherever you host — Netlify, Vercel, Cloudflare Pages, or your own server.
How fast will I rank?›
No promises — anyone who promises is selling you something else. Getting listed takes days to weeks. The skill tells you to check Google's own reporting tool at about two weeks and not panic before then. What it does control is that every page goes after a search you have a real shot at.