Claude for SEO: field notes for audits, briefs, and content ops

Claude for SEO works best as a job runner - audits, briefs, and content ops - not as a one-shot optimize-this-page trick and not as a marketplace install list. Load brand + SERP + crawl/GSC inputs, run a fixed workflow, paste prompts (or skills when the loop is stable), and keep humans on claims and publish. For what SEO skills to install first, use What Claude SEO skills to install first. Cluster entry: Claude for marketing; day-to-day loop parallel: How to use Claude for marketing.

When to use Claude for SEO / when not

Use Claude when the job has evidence you can paste: GSC or crawl notes, SERP observations, a brand pack, a URL list, or a draft with a claims list. Strong fits: triage technical/on-page issues into a prioritized table; turn keyword + SERP notes into a page brief; run content-ops QA before publish; spot cannibalization from a query-page map you already built.

Don't use Claude to invent search volume, KD, or rankings without Labs or a connected tool export. Don't ask it for strategy or taste calls with empty site context. Don't run mass thin content farms. Day zero with no URL, no SERP notes, and no brand rules is a setup failure - not a model failure. And don't start by installing 40 SEO skills; that path belongs to install-order discipline on the SEO skills install-first guide, not this pillar. If you only need which three skills first, stop there - this page assumes you already know the jobs and want the workflow.

SEO jobs map - audits, briefs, content ops

Catalog work by job, not by marketplace SKU names. Marketplace titles change; verify same-day if you name a plugin. The table below is an operator decision aid: inputs you must bring, output you should expect, and when a prompt earns a skill. Deeper skills tooling catalog is planned separately (claude-seo-skills); install order stays on the companion B guide. Do not treat row labels as product SKUs.

Keep the map short. If a job isn't weekly, leave it as a chat prompt. Cannibalization and refresh rows are optional - add them only when you already maintain a query-page map or period GSC export; Claude will not invent those datasets.

Workflow - run SEO jobs in 6-8 steps

1. Inventory weekly SEO jobs. Name them: audit triage, page brief, ops QA. If you cannot name one job, you are not ready to prompt.

2. Gather inputs. Brand pack, SERP notes, crawl or GSC exports, target URL/query. Missing critical inputs: mark [GAP]; do not invent fills.

3. Pick the smallest reliable surface. Chat for spikes; Projects for brand + SERP + brief loops; Claude Code when files/commands/bulk matter; skills after the loop is boring; MCP only as a verified data pipe.

4. Run audit triage or page brief. Use the prompt cards below. One job per run.

5. Run content-ops QA before publish. Draft + forbidden claims + live internal-link targets.

6. Human review. Rankings claims, legal/brand, link relevance. Label output draft until a human approves.

7. Graduate stable loops. When the same spec barely changes for several real runs, thin SKILL.md or install a matching skill after same-day name check. Order: install-first.

8. Prune. Unused skills still cost listing budget; keep the anti-bloat spirit from the install-first guide.

Same discipline as the Projects-first marketing loop: brief before draft, QA before ship.

Prompts (>=3 copy-paste)

Three job-tied specs. Prompt discipline pattern: Claude prompts for marketing.

1. SEO audit triage

Role: SEO analyst. Inputs: URL, crawl/GSC notes, goal, constraints. Output: CRITICAL/HIGH/MEDIUM table with evidence, owner, failure test. When not: inventing crawl data; full site migrations.

You are an SEO analyst. Triage issues from PROVIDED evidence only.
REQUIRED INPUTS: Target URL(s); Crawl or GSC notes (paste); Business/SEO goal (one sentence); Constraints (or "none").
STEPS: Restate job; extract issues only from pasted notes; table severity|issue|evidence|impact|owner|failure test; top 5 actions; mark [GAP].
RULES: Never invent search_volume, KD, rankings, index status, or marketplace skill names.
OUTPUT: severity table | top 5 actions | [GAP] list | QA self-check.

2. Page brief from intent

Role: SEO content strategist. Inputs: primary KW, SERP notes, ICP, URL, must-not. Output: H1 options, outline, must-cover / must-not, internal-link targets. When not: empty SERP notes; ranking guarantees.

You are an SEO content strategist. Write a page brief from intent + SERP notes only.
REQUIRED INPUTS: Primary keyword; SERP notes (paste); ICP; Target URL or "new"; Brand voice; Must-not claims; Live internal-link candidates (or "none").
STEPS: Intent line; three H1 options; H2/H3 outline with must-cover entities; must-cover vs must-not; links only from candidate list; [GAP] questions.
RULES: No promised rankings/traffic lifts; no invented competitor metrics; no marketplace skill dump.
OUTPUT: intent | H1 options | outline | must-cover/must-not | link targets | [GAP].

3. Content-ops QA gate

Role: SEO editor. Inputs: draft, forbidden claims, live /guides/ targets, primary KW. Output: pass/fail checklist + rewrite bullets. When not: legal sign-off; inventing citations.

You are an SEO editor running a pre-publish QA gate.
REQUIRED INPUTS: Draft (paste); Primary keyword; Forbidden claims; Live internal-link targets; Brand voice do/don't (optional).
STEPS: Intent match; checklist (title/H1, thin sections, hallucinated metrics, unsupported citations, live links only, CTA); pass/fail with evidence; max 8 rewrite bullets; flag invented URLs.
RULES: Never invent search metrics. Never label "ready to publish" - label draft.
OUTPUT: pass/fail table | rewrite bullets | residual risks | human gate reminder.

Agent block - SEO field-notes operator

Paste as Project instructions (or Claude Code agent preamble) after brand + SERP files are loaded. Enforces job match, inputs, and anti-hallucination.

You are the SEO field-notes operator (SEO ops runner) for this Project.

MATCH: Classify each request as exactly one job - audit | brief | ops QA. If unclear, ask once. If the user wants "what SEO skills to install first," point them to /guides/what-claude-seo-skills-to-install-first and stop cataloging installs.

INPUTS: Require URL/query, SERP or crawl/GSC notes, and brand constraints for the job - or mark [GAP]. Refuse empty critical fields.

HARD REFUSALS: Never invent search_volume, KD, rankings, CTR curves, or traffic lifts. Never invent marketplace skill names or plugin SKUs. Never invent live URLs or ship dead /guides/ links. Never label output "approved" or "ready to publish."

OUTPUT every run: (1) recommendation for the matched job, (2) next action, (3) filled QA checklist, (4) open questions only - no invented answers.

If the job is Claude Code-deep SEO automation or GEO/AEO systems, say those spokes are planned and stay on audit/brief/ops here.

Skills path / stub

When a weekly brief or QA loop stops changing, package it as a thin skill - don't start with a mega pack.

Stub pattern: a thin SKILL.md named seo-page-brief with frontmatter trigger, required inputs, output schema, and hard refusals. Honest outline only; not a downloadable zip.

Surfaces (verify same-day): Projects custom skills · Claude Code .claude/skills/ · plugins/marketplace after name check. Anthropic describes Agent Skills as reusable folders with SKILL.md that load on demand when the description matches the request - custom skills work across Claude Code, claude.ai upload, and the API (Agent Skills overview, Checked 2026-09-15).

Live bridges: What Claude SEO skills to install first · How do I add Anthropic skills marketplace? · How do I use Claude Code skill-doctor?. Marketing twin pattern (do not duplicate an SEO catalog here): Claude marketing skills. A fuller SEO skills tooling catalog is planned (claude-seo-skills); Claude Code-for-SEO depth and GEO/AEO are also planned - no dead links from this draft.

Stack glance

Smallest reliable system wins. MCP without a human check on exported numbers is still a hallucination risk.

Pitfalls + quality bar

• Hallucinated volumes/KD without Labs or connected tool exports.
• Autopilot publish / thin AI content farms - first pass is draft.
• Skill bloat / overlapping triggers - prune; use skill-doctor when listings get noisy (skill-doctor).
• Plugin READMEs as strategy - a command list is not a content brief.
• Dead internal links - only ship live /guides/ paths; mark planned spokes in prose, don't link them.
• Scope creep in one chat - audit to full brief to publishable article is three runs, not one thread.

Quality bar: evidence fields · no invented metrics · on-brand · human QA for customer-facing · live internal links only. If a claim lacks a source in your inputs, bracket it or delete it before publish.

Light GEO note: answer-first extractability and clear tables help citation-style surfaces, but optimize-for-AI-overviews is still SEO - keep claims tied to your inputs. Deep GEO/AEO belongs on a planned spoke, not this pillar.

FAQ

What is Claude for SEO?

A way to run audits, briefs, and content ops with fixed inputs and human QA - not a single product SKU named Claude SEO, and not a marketplace dump.

Claude chat vs Claude Code for SEO?

Chat (or Projects) for one-offs and brief loops. Claude Code when files, commands, or bulk repo work matter. Deep Claude Code-for-SEO is planned separately.

Do I need SEO skills installed first?

Usually no. Stabilize prompts in a Project first, then graduate. Install order and anti-bloat: What Claude SEO skills to install first.

Can Claude replace an SEO specialist?

No. It accelerates analysis and drafts. Judgment, accountability, and publish decisions stay human.

How is this different from Claude SEO skills guides?

This page = jobs and workflows (audits, briefs, ops). The install-first guide = what to install first. A skills tooling catalog is planned later - do not expect that inventory here.

Where to go next

Run one page brief today with real SERP notes. Keep a human on publish. Map: Claude for marketing. Loop pattern: How to use Claude for marketing. Prompt discipline: Claude prompts for marketing. Skills install order: What Claude SEO skills to install first. Marketplace path: How do I add Anthropic skills marketplace?. GitHub SEO skill install: How do I find and install a Claude SEO skill from GitHub?. Optional compare: Claude vs ChatGPT for marketing. Jobs beat plugins; inputs required; prompts then skills.