Build SEO pages from real conversations
Evidence, not vibes: real search data decides what to build, verified user quotes make it impossible to copy. The open-source seo-page-builder methodology, condensed.
Set it up
1. Connect the Octolens MCP — add the server URL in your client and sign in. OAuth, no API key. Per-client instructions · docs
https://app.octolens.com/api/mcp/v22. Hand your agent the playbook — hit Copy below (it grabs the instruction plus the full playbook), fill in the placeholders, and paste it into your agent. Pasting the playbook instead of describing the idea makes the agent run this exact workflow rather than improvise its own.
Run the playbook below once. Topic: "{TARGET_TOPIC}". Site: {SITE_URL}. Output: {OUTPUT_FILE_PATH}.What it does
- ✓Scopes the page type first — it decides what gets mined: complaints, recommendations, or head-to-head stories
- ✓Pulls real keyword numbers and reads the live SERP like an editor: format, list length, PAA questions, weaknesses to beat
- ✓Mines 6–10 verbatim, URL-verified user quotes — the content competitors can't copy and LLMs can't hallucinate
- ✓Verifies every price and claim against the official source on the day of writing
- ✓Writes honestly: competitor strengths credited, own cons included, vendor affiliation disclosed
- ✓Delivers a draft with schema and a research summary — never publishes
Delivers: A ready-to-review page draft built on the seo-page-builder pipeline: SERP-matched format, verified quotes with links, day-fresh pricing, honest comparisons, and JSON-LD.
Inputs
| Placeholder | Required | Description | Example |
|---|---|---|---|
| {TARGET_TOPIC} | Required | The topic or search query the page should rank for | datadog alternatives |
| {SITE_URL} | Required | Your site — used for voice, cannibalization checks, and disclosure | https://acme.dev |
| {OUTPUT_FILE_PATH} | Required | Where the agent writes the draft | ./drafts/datadog-alternatives.md |
How your agent runs it
- 1
Scope the page — Topic + page type (roundup, listicle, vs, pricing guide, how-to) — the type decides what to mine.
- 2
Keyword research with real numbers — Volumes and difficulty from your SEO tool (never guessed), then the live SERP: winning format, PAA questions, cannibalization check.
seo_toolweb_fetch - 3
Mine primary sources — 6–10 verbatim quotes matched to the page type, deduped and URL-verified; the dominant pain becomes the opening hook.
list_keywordslist_mentions - 4
Verify every fact — Current pricing and tier gates from official pages, fetched the day of writing.
web_fetch - 5
Write the page — SERP-matched structure, quote wall early, honest pros/cons and decision framework, FAQ from real PAA questions, JSON-LD.
- 6
QA and hand off — Quote URLs, prices, schema, and honesty checked — then a draft plus research summary, never a publish.
write_file
Tools used
list_keywordsOctolens MCPFind tracked topic, category, and competitor keywordslist_mentionsOctolens MCPMine 12 months of mentions for verbatim, linkable quotesseo_toolYour MCP clientAhrefs / DataForSEO MCP for volumes, difficulty, and SERP dataweb_fetchYour MCP clientRead the SERP, ranking pages, and official pricing pageswrite_fileYour MCP clientWrite the markdown draft to your project
All 23 Octolens MCP tools are documented on the MCP page and in the docs.
The full playbook
Replace the placeholders, then paste this into your agent.
Build one page, fully researched — never batch-generate. This is the
condensed version of our open-source seo-page-builder skill
(github.com/octolens/seo-page-builder); install that for the full
methodology.
Hard rules, non-negotiable:
- Real quotes only: verbatim, from public posts, with working links.
Never fabricate, never paraphrase-as-quote. Verify every URL resolves.
- Verify every fact at write time — pricing and plan limits from the
official page, fetched today. Stale pricing is the #1 defect.
- Honesty ranks: credit competitors' genuine strengths, include your own
product's cons, disclose vendor affiliation on the page.
- Draft for review. Never publish.
Phase 0 — Scope:
The page targets "{TARGET_TOPIC}" on {SITE_URL}. Determine the page type
(alternatives roundup / best-of listicle / vs comparison / pricing guide /
how-to) — it decides what to mine: complaints for alternatives pages,
recommendations for listicles, head-to-head experiences for vs pages,
"is it worth it" threads for pricing guides.
Phase 1 — Keyword research with real numbers:
If an SEO tool is connected (Ahrefs, DataForSEO), pull volume, difficulty,
and CPC for the modifier family: [topic], best [topic], [topic] tools,
[topic] pricing, question forms. Never guess numbers — if no tool is
available, ask the user for them. Then fetch the live SERP for the primary
keyword and read it like an editor: what format ranks, how long the
ranking lists are, what the People Also Ask questions are (they become
the FAQ), and where the ranking pages are weak (stale dates, no pricing,
no primary sources). If {SITE_URL} already has a nearby page, flag
possible cannibalization and recommend refresh-vs-new before continuing.
Phase 2 — Primary-source mining (the moat):
Use Octolens to collect 6–10 verbatim quotes matched to the page type:
call list_keywords, then list_mentions across topic, category, and
competitor keywords (last 12 months, sorted by relevance). Select quotes
that name specific pain — price figures, missing features, wasted time —
in searcher language. Dedupe by author. Verify every URL is live. The
dominant pattern across quotes (e.g. "pricing requires a sales call")
becomes the page's opening hook.
Phase 3 — Fact verification:
For every product on the page, fetch the current official pricing page.
Record entry price, tier gates, API access. One figure used consistently.
Phase 4 — Write:
Match the SERP-winning format. Open with the Phase 2 hook and a citable
2–3 sentence direct answer. Put the quote wall early — each quote linked,
attributed. Roundups get a comparison table plus per-tool sections with
verified pricing, pros/cons, and "Best for". Add a decision framework
that recommends competitors where they honestly win, an FAQ built from
the real People Also Ask questions, and JSON-LD (Article always; ItemList
for roundups; FAQPage matching the rendered FAQs exactly).
Phase 5 — QA and handoff:
Check: every quote URL opens a real post, all prices verified today,
schema matches rendered content, competitor strengths and own-product
cons present. Write the draft to {OUTPUT_FILE_PATH} and post a summary:
target keyword and volume, format decision and why, the quotes used, and
anything stale you discovered on the rest of the site. Do not publish.More skills
Grow your AI search visibility (GEO)
Find the Reddit threads, listicles, and comparisons AI engines cite where competitors appear and you don't — with drafted contributions for each gap.
Write copy in real customer language
Mine six months of mentions for the exact words customers use to describe you, then rewrite your page copy from that vocabulary — with receipts.
Turn mentions into Linear issues
Catch bug reports and feature requests posted in public — on Reddit, X, LinkedIn, Hacker News, GitHub — and file them as Linear issues automatically.
Give your agents context on what happens online
API, webhooks, and MCP on every plan. If you're paying us anything, you own your data.
No credit card required · API access on all plans · Cancel anytime