agentleFS
alex-voloshin-devalex-voloshin-dev/ai-skills/.agents/skills/geo-writer/SKILL.md

geo-writer

Skill2 starsChanged 6 months agoMITView on GitHub

Optimize any text for GEO/AEO — Generative Engine Optimization and Answer Engine Optimization. Use when writing or editing public-facing text intended to be cited or quoted by LLMs (ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews) — blog posts, landing pages, help articles, glossary entries, FAQ blocks, programmatic SEO templates, pillar guides. Applies macro/meso/micro structure rules, answer-first sections, entity-first phrasing, schema requirements, freshness cadence, and platform-specific citation patterns.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do NOT apply to: internal docs, ADRs, code comments, commit messages, AI asset files.

    line 21
  2. Distinctive:

    Verifiable > persuasive: Every non-trivial claim needs a statistic, date, source, or named example.

    line 29
  3. Very specific:

    Do: ## What is an AI-Readiness Score?

    line 54
  4. Very specific:

    Do not: ## Understanding the Score → first sentence says When it comes to AI search, there are many things to consider....

    line 56
  5. Very specific:

    If marketing/MARKETING.md is absent — ask the user for canonical brand form, category name, and metric names before writing.

    line 103
  6. Very specific:

    AI-generated filler: Always run @humanizer before publishing.

    line 139
  7. Distinctive:

    When producing GEO-optimized text, deliver:

    line 144

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

GEO Writer

Makes public-facing text extractable and citable by LLM-based search engines. Classic SEO is losing ~25% of traffic to AI Overviews, ChatGPT, Perplexity, Gemini, and Claude — content must be structured for extraction, not just for reading.

Based on the GEO-SFE framework (arXiv 2603.29979) and 2026 practitioner sources. Full guide in geo-writing-guide.md. Pre-publish checklist in pre-publish-checklist.md.

When to Use

  • Blog posts, articles, pillar guides
  • Landing pages, help articles, glossary entries
  • FAQ blocks, comparison pages, tutorials
  • Programmatic SEO templates
  • Any text meant to be surfaced or cited by an AI engine

Do NOT apply to: internal docs, ADRs, code comments, commit messages, AI asset files.

Core Principles

  1. Write for extraction, not just reading. LLMs lift 30-180 word chunks. Every section must stand alone.
  2. Answer first, context second. The first sentence of each H2/H3 fully answers the implied question in 30-60 words.
  3. One topic per section. One H2 = one entity or question. Do not merge sub-topics.
  4. Entity-first, not keyword-first. Name the product, brand, person, or concept explicitly. Never start paragraphs with It, This, They referring to prior sections.
  5. Verifiable > persuasive. Every non-trivial claim needs a statistic, date, source, or named example.

Process

  1. Read project brand context — marketing/MARKETING.md at project root (if exists) for canonical brand terms, product name, category label, metric names, audience. Use these verbatim.
  2. Draft macro structure — single literal H1, 5-9 H2 sections each phrased as a standalone question or declarative topic, TL;DR section at end.
  3. Apply meso chunking — paragraphs 40-80 words (hard cap 120), sections 120-180 words, one idea per paragraph, statistic/named source every 150-200 words.
  4. Apply micro emphasis — bold 3-5 most quotable phrases per section, bullet lists for 3+ items, tables for any 2+ entity comparison across 2+ attributes.
  5. Insert high-leverage formats — FAQ block (4-8 Q&A), comparison table where natural, named stats with year, direct quotes from named experts.
  6. Attach schema — Organization + Article + Person on every article; FAQPage for any Q&A block; HowTo for tutorials; Product/SoftwareApplication for tool pages; DefinedTerm for glossary.
  7. Run @humanizer — remove AI-sounding patterns. GEO and humanizer are complementary: GEO optimizes structure, humanizer optimizes voice.
  8. Run pre-publish checklist — see pre-publish-checklist.md. Every item must pass or be explicitly waived.

Answer-First Template

## <Question phrased as H2>

<30-60 word direct answer, no preamble, entity named explicitly.>

<Optional 1-2 supporting paragraphs, 40-80 words each, with at least one number or named source.>

<Optional bullet list, table, or code block.>

Do: ## What is an AI-Readiness Score? → first sentence fully defines the term, names the product, lists the dimensions.

Do not: ## Understanding the Score → first sentence says When it comes to AI search, there are many things to consider....

Format Patterns That Get Cited

Ranked by observed citation lift:

PatternWhy it worksWhen to use
Comparison tablesClean extraction across rows/columnsTools, pricing, competitors, features
FAQ + matching JSON-LD3.1-3.2× higher extraction vs proseEnd of every pillar page
How-to numbered stepsMaps to HowTo schemaTutorials, onboarding, fix-it posts
Definition + example pairsMatches X is Y retrieval patternGlossary, concept explainers
Named stats with year+22% AI visibility in 2026 is quotableIntros, thought leadership
Direct quotes from named expertsQuotation patterns boost visibility ~37%Trend posts, research summaries

Comparison articles lead citation rates at ~32.5% across all formats.

Schema / JSON-LD (Mandatory)

On every page

  • Organization (homepage + sameAs on every article)
  • WebSite with potentialAction for sitelinks search
  • BreadcrumbList on every non-homepage URL

By content type

Content typeRequired schemaNotes
Article / blogArticle + Person (author)datePublished, dateModified, author.sameAs
Pillar / guide with Q&AArticle + FAQPageQuestions must literally match prompt phrasing
Tutorial / walkthroughHowTostep array with 3-8 steps
Product / tool pageProduct or SoftwareApplicationaggregateRating if honest data exists
Glossary entryDefinedTerm in DefinedTermSet

FAQPage is the single highest-leverage schema: 4-8 questions per page, 8-20 words each; answers 40-80 words, self-contained, matching the prompt phrasing; never duplicate marketing copy. Validate in Google Rich Results Test and schema.org/validator.

Entities and Brand Signals

Off-site brand mentions correlate with AI visibility at 0.67 — stronger than any on-site factor. Mentions beat backlinks 3:1 for AI Overviews.

Rules:

  • Use the canonical brand form verbatim — from marketing/MARKETING.md. No variant spellings, no acronyms mid-article.
  • Use the canonical category name and metric names — invented variants fragment the entity graph.
  • Link first mention of any external entity to its canonical page (Wikipedia, official docs).
  • Internal links — every pillar page must interlink to glossary terms and at least 3 related articles.
  • Never change terminology mid-article. GEO score in one paragraph and AI-readiness rating in the next confuses retrieval.

If marketing/MARKETING.md is absent — ask the user for canonical brand form, category name, and metric names before writing.

Freshness Rules

Pages not refreshed quarterly lose AI citations at ~3× the normal rate.

Content typeRefresh cadence
Time-sensitive (algo updates, model releases)24-48 hours
FAQ on pillar pagesMonthly
Evergreen guidesQuarterly
GlossaryQuarterly
Landing / commercial pagesQuarterly

On every refresh: update dateModified in schema, add visible Updated on YYYY-MM-DD line under H1, replace at least one stat or example with a fresher one. Perplexity materially favors updates within the last 2-3 days.

Platform Differentiation

EngineDominant sourceOptimization lever
ChatGPTWikipedia (~48%), training dataSemantic authority, long-lived authoritative pages, Wikipedia entity parity
PerplexityReddit (~47%) + live webFreshness (2-3 days), Reddit presence, clear dates
Google AI OverviewsOrganic top 10 (99%)Classic SEO + schema; top-10 or not cited
GeminiYouTube + Knowledge GraphVideo content with chapters, Knowledge Panel
ClaudeCurated authoritative sourcesCite your own sources inline; neutral, evidence-based tone

Off-site portfolio for pillar topics: 1 Reddit post/comment in a relevant subreddit, 1 YouTube video with descriptive chapters (YouTube overtook Reddit as the #1 LLM citation source in October 2025 — ~16% vs ~10%), 1 LinkedIn long-form post for B2B visibility.

Anti-Patterns

  • Burying the lede. In this post we will explore... openings.
  • Pronoun drift. Paragraphs starting with It, This, They referring to earlier entities.
  • Walls of text. Paragraphs over 120 words.
  • Opinion without a named source. Experts say is invisible to RAG.
  • Duplicate FAQ between body prose and FAQPage schema. LLMs de-duplicate aggressively.
  • Undated statistics. Studies show +30% with no year or link.
  • AI-generated filler. Always run @humanizer before publishing.
  • Changing terminology mid-article. Confuses the entity graph.

Output Format

When producing GEO-optimized text, deliver:

  1. Draft — macro/meso/micro structure applied
  2. Checklist pass — run pre-publish-checklist.md, report each item
  3. Schema block — JSON-LD for the content type, ready to embed
  4. Humanizer pass — applied before final delivery
  5. Follow-ups — off-site companion assets suggested (Reddit / YouTube / LinkedIn), freshness cadence noted

Integration

  • Rule: geo-content rule auto-activates this skill for public-facing content meant to be cited by AI engines
  • Companion skill: @humanizer — GEO structures extraction, humanizer removes AI voice. Always run both.
  • Used by workflows: /blog-post (step 3.5), /marketing (content operations), /docs (public-facing content), /seo-review (GEO audit alongside classic SEO)
  • Used by skills: content-creation (structural gate), marketing-operations (blog and landing operations), social-media-manager (LLM-quotable post framing)
  • Companion resources: geo-writing-guide.md (full reference), pre-publish-checklist.md (PR checklist template)
  • Brand context source: marketing/MARKETING.md at project root — canonical terms, audience, category

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.