SEO / AI Search Audit as a Product

Credit — Nathan Gotch

Source thread by Nathan Gotch (@nathangotch) — SEO + co-founder (bio links his product/company on X).

Original post: x.com/i/status/2083204242383274459 (31 Jul 2026) — “Treat your SEO / AI search audits like a product…”

Pricing he states: they sell these audits for $10,000; he suggests others can sell in the $2,500–$5,000 range even without massive experience — if the report is actionable, clean, and brand-personalized.

Process steps below are his; Atlas / Shadstone evaluation and stack mapping are ours.

Core idea

Treat SEO / AI search audits like a product. It’s never been easier to create actionable (but clean) reports — custom webpages, personalized to the brand, doable even without deep coding experience.

Not a 80-page PDF graveyard. A living, presentable deliverable grounded in a client knowledge base, prioritized with ICE, human-reviewed twice, then shipped as a password-protected page.

Part 1 — Setup

1. Source of truth on GitHub — repo per client (or client folder) for artifacts + findings.
2. Install Codex or Claude — coding agent as the production engine.
3. Connect Rankability MCP — @getrankability (his stack).

For us: same pattern as RankingSolution brain via MCP — agent talks to a brain, not raw DataForSEO only. See also CLI vs MCP: keep thin MCP for the brain; own report/build scripts in the repo.

Part 2 — Nathan’s 15-step audit factory

4. Crawl + knowledge base (AI grounding)

Crawl the website (+ external sources) into a client knowledge base. Bare minimum artifacts:

Maps cleanly to our five-note vault (offer, client, voice ≈ customer-language, playbooks, rulings) plus an explicit entity record for SEO/AEO.

5–8. Research & audits

5. Pick 3 critical categories to dominate → keyword research on those only (focus beats spray).
6. Connect GSC + GA + Bing Webmaster → technical + content audit.
7. Page-level audits (not only site-wide).
8. Off-site opportunities: brand inconsistencies vs entity record, backlink gaps, review gaps.

9–13. Synthesize & write

9. Dump findings into a folder; give the agent access.
10. Agent synthesizes and builds a priority list with ICE (Impact, Confidence, Ease).
11. 🛑 Human review
12. Formulate final written report.
13. Collect screenshots. Optional: Claude Design (or similar) for custom visual assets.

14–15. Ship as a product page

14. Agent builds a dedicated, password-protected webpage (personalized to brand — not a generic PDF template dump).
15. 🛑 Human review again before client delivery.

His tool stack (named)

• Rankability MCP (@getrankability)
• Codex (or Claude)
• Google Search Console
• Google Analytics
• Bing Webmaster Tools

Why this matters for Shadstone Atlas / RankingSolution

We already invest in company / SEO brains (Atlas internally; RankingSolution for portfolio SEO; MCP so agents don’t hit raw DataForSEO every time — see backlinks guide). Nathan’s post is a blueprint for turning that brain into a sellable, repeatable audit product — not only an internal dashboard.

Team evaluation question: Does Atlas help us run this factory — or only store metrics?

Atlas feature checklist (evaluate against this workflow)

Suggested Atlas modules (names for the team to argue about)

1. Client Grounding Pack — the five md artifacts + entity sign-off
2. Audit Run — crawl + GSC/GA/Bing + page audits + off-site
3. ICE Backlog — prioritized actions
4. Report Site — password page / branded microsite
5. Agent Session Bridge — MCP + optional CLI pulse export for deep work

How we’d run this with our stack (today)

1. Client (or brand) folder in GitHub / HEOS-style vault
2. Grounding artifacts (entity-record.md first class)
3. RankingSolution / Atlas brain for data + history
4. Thin MCP for agent access OR CLI scripts for heavy pulse
5. Agent synthesis → ICE list
6. Human review (SEO lead)
7. Generate password report page (Astro/static or Atlas export)
8. Human review → deliver
9. Loop: findings feed brain so next audit compounds

That last step is our edge if Atlas is a real brain — Nathan’s product is the audit; our software should make the second audit cheaper and smarter.

Team evaluation questions (Shadstone Atlas)

1. Can we export a full “Nathan pack” (5 grounding files + findings + ICE) in one click?
2. Is entity-record first-class, or buried in notes?
3. Can an agent open an audit run end-to-end without leaving the brain for secrets?
4. Do we have a report webpage generator, or only PDF/dashboard screenshots?
5. Where do human review gates live — and are they enforced?
6. What’s our SKU: $2.5k / $5k / $10k — and which Atlas features are required for each?
7. How does this join the company-pulse CLI direction for multi-source data?

Related on this site

References

• Nathan Gotch — @nathangotch · SEO/AI audit as product thread
• Rankability — @getrankability (as cited in his stack)