Advanced SEO Prompting: The “Armored MASTER PROMPT” Decoded (Checklists)

TL;DR: An “armored” master prompt = clear role + measurable objectives + rich context + SEO constraints + strict format + anti-hallucination safeguards + acceptance criteria. We industrialize with variables, checklists, and output tests (semantic QA, internal linking, tagging). Below: ready-to-use templates, control matrices, and audit plans for WordPress.

Definition & Principles

Advanced SEO prompting consists of designing robust instructions that enable a language model to produce SEO deliverables that are reliable, measurable, and publishable (briefs, HTML structures, content, JSON-LD schemas, semantic clusters, internal linking ideas). We speak of a master prompt when the methodology, requirements, and editorial policy are encapsulated in a single reusable artifact.

Risk #1: hallucination (inaccurate facts, invented sources). Solution: safeguards in the prompt + systematic verification + obligation to cite verifiable sources.

The “armoring” strengthens these aspects through: (1) variables (keywords, audience, tone, target SERP), (2) constraints (EEAT, internal linking, format), (3) tests (checklists, rejection rules if info is missing), (4) a legal framework (copyright, citation, GDPR), (5) modes (draft, expert, synthesis).

Anatomy of the “Armored MASTER PROMPT”

Here is the structure I use on assignments. It adapts to ChatGPT, Claude, or any LLM compatible with system prompts and formatted outputs.

1. Role & mission: “You are a senior SEO consultant and web writer expert in popularization.”
2. SMART objectives: concrete KPIs (traffic, ranking, click-through rate, crawl depth, indexing rate).
3. Context & input data: main keyword, intent (informational / transactional), SERP features, competitors, existing assets.
4. SEO constraints: EEAT (expertise, sources), tagging (H1-H3), natural density, entities, internal linking, meta, canonical.
5. Style & brand voice: tone, technical level, allowed/forbidden jargon, languages and regional variants.
6. Expected format: WordPress-ready HTML, JSON-LD, tables, <aside>/<blockquote> tags, target length.
7. Safeguards: prohibitions (inventing prices/dates), source policy, refusal instructions (“if info is missing, ask for it or propose a flagged partial version”).
8. Acceptance criteria: measurable checklists (10+ H2/H3, 2 internal links, 6 reliable sources, 4 unique images, etc.).
9. Multiple outputs: (a) brief, (b) outline, (c) article, (d) meta, (e) FAQ, (f) schema.
10. Post-processing: self-check (proofreading, anti-plagiarism, fact-check), A/B test suggestions, dev/UX to-do.

Settings: keep explicit variables (e.g. {{mot_cle_principal}}, {{persona}}, {{objectifs}}) to switch from one case to another without rewriting the prompt.

Checklists (before / during / after)

Before (pre-prompt)

• Define the keyword intent (info / nav / transactional).
• List 3–5 competitors and their angles (SERP page 1).
• Collect entities & questions (People Also Ask, autosuggest, forums).
• Choose the format: article, guide, category page, product sheet, comparison.
• Set the KPI: target position, CTR, outline depth.
• Validate the brand voice and forbidden elements.

During (in the prompt)

• Require external sources verifiable (3–6) + 2 internal links.
• Enforce a format (valid HTML, JSON-LD, meta lengths).
• Include acceptance criteria checkable by the AI itself.
• Define limits: “do not invent”, “if uncertain, say so”.
• Add a QA section (self-check) at the end of the output.

After (post-prompt / QA)

• Check SERP alignment (H2 titles vs competitors, differentiating angle).
• Test internal linking (2–4 internal links + varied anchors).
• Verify sources (accessible links, no paywall if forbidden).
• Proofread EEAT (author, expertise, disclaimers if needed).
• Validate schemas (schema.org validator / Rich Results Test).

SEO Prompt Templates (copy-paste)

Tip: store these templates in a shared folder. Replace the variables between {{ }} and keep the same structure for the whole team.

1) MASTER prompt hardened — WordPress EEAT Article

ROLE: You are a senior SEO consultant and expert web writer.
OBJECTIVE: Write a WordPress article ranking for {{mot_cle_principal}} in {{langue}} for {{persona}}, EEAT compliant.

CONTEXT:
- Intent: {{intention}}; Country/language: {{pays_langue}}; Level: {{niveau}}
- SERP competitors (summarize each angle): {{concurrents}}
- Entities to cover: {{entites}}
- Available internal links: {{liens_internes}}

SEO CONSTRAINTS: – Length ≥ {{mots_min}} words; H1 contains {{mot_cle_principal}}; 10+ H2/H3; 2+ internal links; 6 reliable external sources; 4 unique images (public URLs). – Meta title (~60c) and description (~150c); JSON-LD Article + FAQ; style: {{ton}}; readability: {{niveau_lecture}}. – Forbidden: no AI images, no factual inventions, no unsourced rates/prices/dates. OUTPUT FORMAT: – Complete WordPress-ready HTML (doctype, head, body), UTF-8 encoding, single block. – Include <nav> summary, <aside>/<blockquote> for tips, table if relevant. – Add FAQ section (≥ 5 Q/A) + sources (6+). SAFEGUARDS: – If information is missing, produce a partial version and clearly indicate assumptions. – If a statement is uncertain, say so and suggest a source to verify. ACCEPTANCE CRITERIA (checklist at end): – [ ] H1 ok; [ ] Meta ok; [ ] 2+ internal links; [ ] 6+ sources; [ ] 4 unique images; [ ] Valid JSON-LD; [ ] Compliant tone. DATA: {{notes_biz}}

2) Prompt “SEO Brief & Plan” (before writing)

Mission: produce an actionable SEO brief for {{mot_cle_principal}}.
Deliverables: intent, differentiating angle, H2/H3 plan, entities, People Also Ask, FAQ, internal linking, title/description (3 variants), EEAT risks.
Constraints: tone {{ton}}, audience {{persona}}, target length {{mots_min}}.
Output: JSON + summary table and 5 image ideas (free sources).

3) Prompt “On-Page Optimization” (existing text)

Mission: optimize the existing page {{url}}.
Deliverables: H1/H2 suggestions, semantic enrichment, length compression, meta rewrite, FAQ schema, 4 relevant internal links with varied anchors.
Constraints: preserve meaning, avoid cannibalization, respect brand voice {{ton}}.
Output: before/after diff + WordPress implementation checklist.

4) Prompt “Semantic Cluster & Internal Linking”

Mission: generate a cluster around {{mot_cle_pilier}}.
Deliverables: 1 pillar page + 6 to 12 satellite pages; for each satellite: keyword, intent, angle, 3 internal links (-> pillar & other satellites).
Constraints: avoid duplication, clear hierarchy (breadcrumbs), avoid cannibalization.
Output: table + link graph (edge list).

5) Prompt “Snippets & SERP Features”

Mission: maximize eligibility for SERP features for {{mot_cle_principal}}.
Deliverables: optimized snippet (40–50 words), ordered list (how-to), comparison table, structured data (FAQPage, HowTo if relevant).
Constraints: clear language, no exaggeration, sources to cite.
Output: ready-to-paste blocks + validatable JSON-LD schemas.

QA & Acceptance Criteria (automatable)

Your master prompt must produce its own QA checklist at the end of the output. Here is a base:

Aspect	Criterion	Quick test
Meta	Title ≤ 60c / Description ≤ 160c	Count characters (spaces included)
Structure	Unique H1, 10+ H2/H3	Regex search <h2>
EEAT	Author + 6 external sources	Presence of <ul> “Sources”
Internal links	≥ 2 internal links	Check href starting with /
Images	≥ 4, unique, descriptive alt	Compare URLs; check alt
Schemas	Valid Article + FAQ JSON-LD	Rich Results Test
Tone	Brand compliant	Human reading + editorial guide

Industrialization workflow (stack & roles)

Roles

• SEO Strategist: chooses targets, sets KPIs, validates angles.
• Writer: applies the master prompt, enriches, clarifies.
• Fact-checker: verifies sources, data, legal compliance.
• WordPress Integrator: tags, schemas, media, internal linking.
• Analyst: monitors performance (impressions, CTR, positions, engagement).

Pipeline

1. Brief & plan (LLM) → human validation (10 min).
2. Writing V1 (LLM) → human enrichment (EEAT, examples).
3. Auto QA (checklist) → human QA (sources, tone, linking).
4. WP integration (HTML + JSON-LD) → publication.
5. Measurement (Search Console, Analytics) → iterations.

Concrete examples

Example A — Technical guides

Problem: very technical articles with risk of errors. Shielding: require official standards (RFC, publisher docs), impose “if uncertain, say: ‘to verify'” and separate the “Sourced facts” section from the rest. Result: fewer post-publication corrections, increased dwell time.

Example B — E-commerce category pages

Problem: generic descriptions. Shielding: variables for purchase criteria, product FAQ, structured comparisons, usage warnings, mention of guarantees/returns. Result: better relevance, improved CTR.

Example C — Pillar page + cluster

Problem: cannibalization. Shielding: linking rules (1 “up” link, 2 “across” links), prohibition to target the same intent, unique angles checklist. Result: better indexing clarity, progress on long-tail queries.

Recommended tools & practices

• Search Console to measure impressions/CTR/positions.
• PageSpeed / Core Web Vitals for page experience (LCP, CLS, INP) — UX influences engagement and indirectly SEO.
• schema.org Validator and Rich Results Test for JSON-LD.
• Editing suites: collaborative editors (comments, change tracking).
• Centralized prompt library (versioned) + nomenclature (e.g. SEO-ART-EEAT-v3).

Recommended Keywords & Semantics

• Main: SEO prompting, master prompt, SEO prompt engineering, AI SEO checklists.
• Long tail: SEO prompt templates, hallucination safeguards, WordPress EEAT prompt, JSON-LD SEO FAQ, automated internal linking.
• Synonyms: prompt recipe, prompt architecture, editorial AI workflow.

FAQ — Advanced SEO Prompting

What is a “fully loaded MASTER PROMPT”? A complete prompt template that includes role, objectives, context, SEO constraints, output format, safeguards, and QA criteria to produce publishable deliverables. How to avoid hallucinations? By requiring citation of verifiable sources, refusal rules (“say I don’t know”), and a final QA that flags any factual uncertainty. Can QA be automated? Partially: meta character count, H2/H3 detection, presence of internal links, JSON-LD validation. Fact-checking remains human for sensitive points. How long should a master prompt be? Long enough to frame (600–1,200 words), but segmented into blocks (role, objectives, constraints) with reusable variables. Should the same prompt be used for everything? No. Create a common base + modules (brief, writing, on-page, clusters) to limit debt and maintain specialization.

Sources

• Google Search Central — Creating helpful and reliable content
• Google — SEO Starter Guide
• Google — Structured Data (documentation)
• Schema.org — Types and properties
• OpenAI — Prompt engineering (best practices)
• Anthropic — Prompting guides
• Nielsen Norman Group — UX writing & readability

---