agentleFS
vidual-labsvidual-labs/digital-marketing-skills/skills/marketing/geo-ai-seo/SKILL.md

geo-ai-seo

Skill3 starsChanged 46 days agoMITView on GitHub

Use when optimizing content to be cited by AI search and answer engines — Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Gemini, Copilot and Claude — through Generative Engine Optimization (GEO). Covers answer-first structure, data anchors, source attribution, entity clarity, FAQ and comparison templates, AI-crawler access and measuring AI referrals. Also use when someone asks about "AEO", "LLM SEO", "showing up in ChatGPT", "AI Overviews" or "why AI doesn't mention our brand". Don't use for paid search keywords (google-ads-keywords), social copy, or landing page conversion (landing-page-funnel).

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 3 are distinctive, and 3 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.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Don't use for: Google Ads keyword work (use google-ads-keywords), social media copy, landing page CRO (use landing-page-funnel), or deep …

    line 29
  2. Distinctive:

    Ask for (or extract from the conversation):

    line 33
  3. Distinctive:

    Entity clarity: Use the full brand/product name on first mention in each section.

    line 77
  4. Very specific:

    Freshness: Update facts, change the visible "Updated" date only when content actually changed, and keep dateModified honest.

    line 79
  5. Very specific:

    AI referrals in GA4: create a channel or segment for sessions whose source contains chatgpt.com, perplexity.ai, copilot.microsoft.com, …

    line 138
  6. Distinctive:

    Ignoring third-party surfaces: If Reddit, YouTube and review sites carry most citations in your category, being absent there caps your GEO …

    line 193

≈ “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

Generative Engine Optimization (GEO)

Overview

GEO is optimization for AI-generated answers. Traditional SEO targets a ranking; GEO targets citation probability — the chance that an LLM-based engine pulls a passage from your page and names you as a source or recommends your brand. The engines read, chunk and extract; the winning content is answer-first, specific, attributed, fresh, crawlable and about clearly identified entities. This skill structures, rewrites and audits content for that, and tells you how to check whether it worked.

When to Use

  • Creating or rewriting pages that should be cited in AI answers
  • Structuring FAQ, how-to, comparison, pricing or "best X" content
  • Auditing why competitors appear in AI answers and you do not
  • Checking technical access for AI crawlers and search bots
  • Setting up measurement for AI referral traffic and brand mentions

Don't use for: Google Ads keyword work (use google-ads-keywords), social media copy, landing page CRO (use landing-page-funnel), or deep schema markup engineering beyond the basics listed here.

Inputs

Ask for (or extract from the conversation):

  • Target queries / questions the content should answer, in the user's words
  • The page or draft (URL or text) and its purpose (inform, compare, sell)
  • Proof assets: original data, customer numbers, studies, expert authors
  • Entity facts: exact brand/product names, category, founding facts, locations, pricing
  • Competitors that currently get cited (if known) or the AI answers the user has seen
  • Access to robots.txt, CDN/WAF settings and analytics, or someone who has it

Data rules: never fabricate statistics, studies or quotes to "add data anchors" — insert [add verified stat + source] placeholders and explain. If you can browse, test the target query in at least two engines and report what they cite; if you cannot, ask the user for screenshots of the answers. Engine behaviour changes monthly; treat the landscape table as a snapshot dated by metadata.updated.

AI Search Landscape (snapshot)

EngineHow it sourcesWhat it rewards
Google AI Overviews / AI ModeGoogle's index; AI Mode runs several sub-queries ("fan-out") and stitches passagesPages already indexed and crawlable, passages that answer one sub-question completely, E-E-A-T signals, freshness
ChatGPT (search)Live web via OAI-SearchBot plus Bing-derived indexClear definitions, step lists, named sources, recently updated pages
PerplexityOwn crawler (PerplexityBot) and live searchDense facts, numbers, recent dates, comparison tables
GeminiGoogle indexSame as AI Overviews; strong on structured, entity-rich content
Microsoft CopilotBing indexStructured Q&A, schema markup, freshness
Claude (web search)Partner search index plus live fetchPrecise, well-attributed passages; readable HTML

Common pattern across engines: a handful of large domains (Reddit, Wikipedia, YouTube, major publishers, review platforms) take a large share of citations. Being present on those — answering in relevant Reddit threads, maintaining accurate Wikipedia/Wikidata facts, publishing on YouTube — is part of GEO, not only your own pages.

Technical Access Checklist

Content cannot be cited if the engine cannot read it.

  • robots.txt allows the search agents you want: OAI-SearchBot, PerplexityBot, ClaudeBot/Claude-SearchBot, Googlebot, Bingbot. Decide separately about training agents (GPTBot, Google-Extended, CCBot, anthropic-ai); blocking training does not have to block search.
  • CDN/WAF (Cloudflare and similar) is not challenging or rate-limiting those agents.
  • No noindex, nosnippet or max-snippet:0 on pages you want cited.
  • Key content is server-rendered HTML, not injected by JavaScript after load.
  • Pages are indexed in Google and Bing (AI Overviews and Copilot only cite indexed pages).
  • Basic schema.org markup present where it fits: Article with author and dateModified, FAQPage, HowTo, Product with offers, Organization with sameAs links to official profiles.
  • llms.txt is an unproven convention; harmless to add, do not expect ranking effects, and Google has said it does not use it.

Content Structure for GEO

  1. Answer first. The first 40–80 words under the title (and under each H2) state the direct answer. No throat-clearing.
  2. One question per section. H2/H3 headings are the questions people ask, phrased naturally. Each section is self-contained so it can be lifted as a passage.
  3. Data anchors. Numbers, dates, ranges, named entities. "Costs 500–2,000 EUR depending on X (as of Sept 2026)" beats "costs vary".
  4. Attribution. "According to [named source], [year]" — link it. Engines weigh sourced claims; unsourced claims look like opinion.
  5. Definition blocks. "X is [one-sentence definition]. [elaboration]."
  6. Lists and tables for procedures and comparisons. Numbered steps are extracted verbatim; tables are parsed as structured evidence.
  7. Entity clarity. Use the full brand/product name on first mention in each section; keep names consistent with your Organization schema and official profiles.
  8. Author and experience signals. Byline with credentials, first-hand observations ("we tested 12 units for 30 days"), an updated date that is true.
  9. Freshness. Update facts, change the visible "Updated" date only when content actually changed, and keep dateModified honest.
  10. Original contribution. Something not on competing pages: your data, a survey, a benchmark, a decision framework. Paraphrases of the top results do not get cited.

Text formation rules

RuleWeakStrong
Answer-first"Many factors influence the cost of X…""X costs 500–2,000 EUR; the main driver is Y. Breakdown:"
Specificity"Many businesses see better ROI""[n]% of surveyed B2B teams reported higher ROI ([source], [year])"
Self-contained step"Then configure it""Step 3: Settings → API keys → Create key with read/write scope"
Entity density"the platform helps with this""[Brand]'s [Product] handles [specific capability] for [segment]"
Definition"There are various views on X""X is [definition]. In practice it means [elaboration]."
No filler"In this article we will explore…"Delete; start with the answer

Section Templates

FAQ block

## What is [topic]?
[Topic] is [one-sentence definition]. [Two sentences of elaboration with one named source.]

## How does [topic] work?
1. [Complete step — who, what, result]
2. [...]
3. [...]

## How much does [thing] cost?
As of [Month Year], [thing] costs:
- [Tier]: [price] — [what is included]
- [Tier]: [price] — [what is included]
Source: [official page], accessed [date].

## What are the best [things] for [use case]?
| Option | Best for | Key fact |
|--------|----------|----------|
| [Name] | [use case] | [specific metric] |

Comparison block

## [A] vs [B]: which is better for [use case]?
| Criterion | [A] | [B] |
|-----------|-----|-----|
| [Metric]  | [value] | [value] |
| [Metric]  | [value] | [value] |

**Verdict:** [A] suits [use case] because [specific reason + data]. [B] wins for [use case] because [specific reason + data].

Brand entity block (About / homepage)

[Brand] is a [category] for [who], founded [year] in [place]. It [one-sentence differentiator with a number]. [Brand] is used by [n] [customers] including [named examples].

Measurement

  • AI referrals in GA4: create a channel or segment for sessions whose source contains chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com, claude.ai. Expect small volumes with high intent.
  • Citation checks: monthly, run the 10–20 target questions in each engine (manually or with a monitoring tool) and log whether the brand is mentioned, cited (linked) or recommended, and which competitor passages are used instead.
  • Google Search Console: AI Overview impressions are included in Search performance; watch impressions-without-clicks on question queries.
  • Brand mention tracking: prompts that describe the category without the brand name ("best [category] for [segment]") test unaided recall in the model.

Output Format

GEO AUDIT / BRIEF: [Page or topic]
Target questions: [list]
Engines checked: [which, date] | Currently cited: [who]

--- TECHNICAL ACCESS ---
[robots.txt / WAF / indexing / rendering / schema findings → fixes]

--- STRUCTURE ---
[Answer-first opening: present / rewrite]
[H2 question mapping: current → recommended]
[Lists / tables: where to add]

--- CONTENT ---
Data anchors to add: [facts needed, with `[add verified stat + source]` placeholders]
Attribution gaps: [claims needing sources]
Entity clarity: [name consistency, schema]
Original contribution: [what unique data/perspective to add]
Freshness: [what to update, how to date it]

--- REWRITTEN SECTIONS ---
[Answer-first opening, FAQ block, comparison block as applicable]

--- MEASUREMENT PLAN ---
[GA4 segment, monthly citation check list, Search Console view]

--- PRIORITY ACTIONS ---
1. [...]
2. [...]
3. [...]

Common Pitfalls

  1. Writing for the model only. Humans still read the page and convert. Precision and readability are compatible; dullness is not required.
  2. Burying the answer. Context-building intros push the extractable passage below the fold of the model's attention. Answer, then explain.
  3. Vague quantifiers. "Many", "several", "a lot" are replaced with numbers or removed.
  4. Fabricated data anchors. An invented statistic that gets cited is a liability. Placeholder and verify.
  5. Blocking the bots you want. Teams block GPTBot for training and accidentally block OAI-SearchBot. Separate the decisions.
  6. Duplicating the top results. Engines favour pages that add something. Bring original data or a distinct framework.
  7. Fake freshness. Bumping the date without changing content is noticed when the facts are stale.
  8. Ignoring third-party surfaces. If Reddit, YouTube and review sites carry most citations in your category, being absent there caps your GEO ceiling.

Verification Checklist

  • Technical access checked: robots.txt agents, WAF, indexing, rendering, snippet directives
  • First 40–80 words of the page and of each H2 give the direct answer
  • Each H2/H3 is a real question or statement users search for
  • Every paragraph names at least one entity (brand, person, place, product)
  • Vague quantifiers replaced; data anchors either verified with sources or marked as placeholders
  • At least one table or numbered list where a comparison or procedure exists
  • Author/experience signals and an honest updated date present
  • One original contribution identified that competitors lack
  • Brand entity facts consistent with schema and official profiles
  • Measurement plan covers AI referrals, citation checks and Search Console
  • Engines and date of the citation check recorded

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.