agentleFS
varunk130varunk130/claude-code-skills/skills/ai-marketing-growth/ai-search-visibility-optimizer/SKILL.md

ai-search-visibility-optimizer

Skill1 starsChanged 22 days agoMITView on GitHub

Produces a Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO) audit: a buyer-intent query set, share-of-answer scores across AI assistants, a citation-source gap map, entity-consistency findings, and a prioritized content and off-site fix plan. Use when a brand is missing or misdescribed in ChatGPT, Claude, Perplexity, or Google AI Overviews answers; organic search traffic is falling while AI referrals rise; competitors are recommended by AI assistants and you are not; or planning content for AI-generated answers rather than blue links.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 9 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    After noticing branded or category demand shifting from search to assistants

    line 25
  2. Distinctive:

    Before a launch or rebrand, to make sure engines describe the new positioning correctly

    line 26
  3. Distinctive:

    When sales hears "the AI told me to look at Competitor A"

    line 27
  4. Distinctive:

    When planning a content roadmap that must earn citations, not only clicks

    line 28
  5. Distinctive:

    Run each prompt at least 3 times per engine (answers are non-deterministic).

    line 46
  6. Distinctive:

    Add a sentiment and accuracy flag per mention: accurate, outdated, wrong, or negative.

    line 55
  7. Distinctive:

    Run every prompt multiple times per engine and report averages, not single screenshots

    line 122
  8. Distinctive:

    Keep a control set of untargeted prompts to separate your impact from engine drift

    line 126
  9. Distinctive:

    Treat a single engine as the whole channel

    line 131

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

AI Search Visibility Optimizer

When buyers ask an assistant instead of a search box, the shortlist is written before they visit your site - this skill gets you onto it.

What this skill is

A repeatable audit-and-fix workflow for how a brand, product, or category appears inside answers generated by Large Language Model (LLM) assistants and AI search surfaces. It measures share of answer on a fixed, intent-weighted query set, traces which third-party sources the engines cite, and converts the gaps into specific content, entity, and off-site actions. It treats AI visibility as a measurable channel with a baseline and a re-test cadence, not a one-time optimization trick.

What it solves

  • Teams optimizing for blue-link rankings while AI answers recommend competitors for the same intent
  • Brand descriptions in AI answers that are outdated, wrong, or borrowed from a competitor's framing
  • One-off "we asked ChatGPT and we showed up" anecdotes with no repeatable measurement
  • Content that ranks but is never cited because it lacks extractable facts, comparisons, or clear entity signals
  • Ignoring the third-party sources (review sites, forums, listicles, documentation) that engines actually cite

When to invoke

  • Quarterly visibility baseline for a product category across AI assistants
  • After noticing branded or category demand shifting from search to assistants
  • Before a launch or rebrand, to make sure engines describe the new positioning correctly
  • When sales hears "the AI told me to look at Competitor A"
  • When planning a content roadmap that must earn citations, not only clicks

Phase 1: Design the query set

Build 30-60 prompts that a real buyer would type, grouped by intent. Fix the set and version it so later runs are comparable.

Intent tierExample prompt patternWeight
Category discovery"Best tools for <job> for a <segment> team"3
Comparison"<Brand> vs <Competitor A> for <use case>"3
Problem-led"How do I <solve pain> without <constraint>"2
Alternatives"Alternatives to <Competitor A>"2
Branded fact"What does <Brand> cost / integrate with / do"1

Rules: include 3-5 persona or segment variants, phrase prompts conversationally, and record engine, model or mode, date, location, and whether browsing or search grounding was on.

Phase 2: Run and score share of answer

Run each prompt at least 3 times per engine (answers are non-deterministic). Score each response:

ScoreMeaning
0Brand absent
1Mentioned in passing or in a long list (position 5+)
2Listed in top 3-4 with neutral description
3Recommended first or as the best fit for the stated need

Add a sentiment and accuracy flag per mention: accurate, outdated, wrong, or negative.

Share of answer (SoA) = sum(score x intent weight) / (max score 3 x sum(intent weight)) across runs
Mention rate = responses with brand / total responses

Report SoA per engine, per intent tier, and versus the top 3 competitors on the same set.

Phase 3: Citation-source gap analysis

For engines that show sources, log every cited Uniform Resource Locator (URL) and classify it.

Source typeExamplesYour presenceCompetitor presenceGap action
OwnedProduct pages, docs, pricing, blogRestructure for extraction
Review and directorySoftware review sites, marketplacesProfile completion, review drive
CommunityForums, Q&A sites, developer threadsAuthentic participation
EditorialListicles, analyst notes, newslettersOutreach with data assets
ReferenceWikis, knowledge bases, open datasetsNeutral factual entries where notable

Rank sources by citation frequency. The top 10 cited domains are the real distribution channel - prioritize gaps there before writing anything new.

Phase 4: Entity and fact consistency check

Engines assemble answers from entity facts. Audit whether the same facts appear consistently everywhere:

FactCanonical valueSiteDocsReview profilesSocial biosStructured data
Category label
One-line description
Pricing model and starting price
Primary integrations
Target segment

Any cell that conflicts is a likely source of wrong or hedged answers. Add Organization, Product, and FAQ structured data (schema.org markup) where appropriate.

Phase 5: Content fixes for extractability

Rewrite or create pages so an engine can lift a correct answer:

  • Answer-first blocks: a 40-80 word direct answer under each question-style heading
  • Comparison pages with honest tables (criteria, where each option fits, where you do not)
  • Facts pages: pricing, limits, security, integrations, stated plainly with dates
  • Original data: benchmarks, surveys, or usage statistics others will cite
  • Specific nouns: name the job, segment, and integration rather than generic claims

Score each priority page 1-5 on: direct answer present, verifiable facts, freshness date, comparison coverage, and crawlability (no critical content behind scripts or logins).

Phase 6: Prioritize and re-test

FixIntent tier affectedExpected SoA lift (L/M/H)Effort (L/M/H)OwnerRe-test date

Re-run the identical query set every 4-8 weeks. Attribute movement cautiously: engines update models and indexes independently of your changes, so track a control set of prompts you did not target.

Output

  • Versioned query set with intent weights and run conditions
  • Share-of-answer scorecard per engine, intent tier, and competitor
  • Accuracy log of wrong or outdated statements about the brand
  • Citation-source map with top 10 cited domains and gap actions
  • Entity-consistency table with conflicts flagged
  • Prioritized fix backlog with owners and a re-test calendar

Operating rules

Always

  • Run every prompt multiple times per engine and report averages, not single screenshots
  • Record engine, mode, date, and grounding settings for every run
  • Benchmark against named competitors on the identical query set
  • Fix factual inconsistencies across owned and third-party profiles before producing new content
  • Keep a control set of untargeted prompts to separate your impact from engine drift

Never

  • Promise rankings or guaranteed inclusion in AI answers
  • Publish fake reviews, astroturfed forum posts, or hidden text aimed at crawlers
  • Treat a single engine as the whole channel
  • Change the query set between runs and then compare scores
  • Optimize branded-fact prompts while ignoring category discovery, where buying decisions start

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.