thesmokedev/geo-skills/geo-platform-optimizer
Claude Code and agent skill by thesmokedev/geo-skills
- community
Checking that the SKILL.md file is still in the repository.
Platform-specific AI search optimization — audit and optimize for Google AI Overviews, ChatGPT, Perplexity, Gemini, and Bing Copilot individually
Install
npx vskill@latest install thesmokedev/geo-skills/geo-platform-optimizerSKILL.md
The same SKILL.md was published earlier as thesmokedev/geo-skills/geo-platform-optimizer.
GEO Platform Optimizer
Core Insight
Only 11% of domains are cited by BOTH ChatGPT and Google AI Overviews for the same query. Each AI search platform uses different indexes, ranking logic, and source preferences. A page optimized for Google AI Overviews may be invisible to ChatGPT, and vice versa. Platform-specific optimization is not optional — it is the foundation of any serious GEO strategy.
How to Use This Skill
- Collect the target URL and the site's primary topic/industry
- Run each platform checklist below against the site
- Score each platform on the 0-100 rubric
- Generate GEO-PLATFORM-OPTIMIZATION.md with per-platform scores, gaps, and action items
Platform 1: Google AI Overviews (AIO)
How AIO Selects Sources
- Only 38% of AIO-cited URLs rank in the organic top 10 — down from 76% (Ahrefs, 863K SERPs / 4M URLs, Mar 2026). 31% of citations come from positions 11-100 and 31% from beyond position 100. The cause is Gemini 3 (Jan 2026) query fan-out pulling from sub-query SERPs — page-3 organic is NOT disqualifying
- Fan-out coverage is the new meta-factor (Zyppy scores it 9.3/10, Jun 2026): engines rewrite one prompt into a cluster of sub-queries and retrieve per sub-query. Win the cluster, not the head term
- AIO strongly favors pages with clean structure, direct answers, and scannable formatting
- Featured snippet optimization has meaningful overlap with AIO optimization
- AIO prefers concise, factual, unambiguous answers — hedging and filler reduce citation probability
- Google AI Mode is a separate surface, not a bigger AIO: AIO↔AI Mode URL overlap is only 13.7%, and AI Mode has 1B MAU (AIO 2.5B) with scarcer citation slots — 4.3 citations per response vs 10.3 in classic search (Shadow, Jul 2026 — ⚠️ secondary source). Optimize for them as distinct targets
2026 Citation Surfaces Beyond the AIO Box
Google's citation real estate now extends beyond AIO and AI Mode themselves:
- Preferred Sources — users can star favorite outlets and get boosted Top Stories placement plus a dedicated "From your sources" section (Google Search blog, Aug 2025, US/India English launch; broader rollout through 2026). This is a Top Stories surface, NOT an AIO mechanism — but for news-adjacent queries it is part of the brand's total Google citation footprint. Earning a user's "preferred" star compounds like a subscription; publishers can link directly to a follow flow.
- "Highly Cited" badges, Community Perspectives, and link carousels — additional in-SERP citation modules attributed to the Google I/O May 2026 announcements. ⚠️ Thin-sourced: this comes from rival skill-pack I/O 2026 notes and we have NOT independently confirmed rollout status or the exact selection mechanics. Treat as watch items — if these modules appear in your SERPs, they are additional extractable surfaces — but do not score them in the rubric until the mechanics are verified.
- Common thread: every new surface still extracts from the same primitives — direct answers, tables, dated authorship, entity clarity. Optimizing the fundamentals covers surfaces that ship faster than any checklist can track.
Optimization Checklist
- Question-Based Headings: Use H2/H3 headings phrased as questions matching real user queries. Check Google's "People Also Ask" for the target topic and mirror those exact phrasings.
- Direct Answer in First Paragraph: After each question heading, provide a clear 1-2 sentence answer immediately. Then expand with supporting detail. The first sentence should be a standalone citation candidate.
- Tables and Structured Comparisons: AIO heavily cites tables. Convert any comparison, pricing, specification, or feature data into HTML tables. Use clear column headers.
- Ordered and Unordered Lists: Step-by-step processes should use ordered lists. Feature lists should use unordered lists. AIO extracts these directly.
- FAQ Sections: Add a dedicated FAQ section with 5-10 real questions. Use proper H3 headings for each question. While FAQPage schema rich results are restricted to govt/health sites since Aug 2023, the content pattern still helps AIO extraction.
- Definitions and Glossary Boxes: For any industry-specific term, provide a clear definition. Format: "[Term] is [concise definition]." AIO frequently cites definitions.
- Statistics with Sources: Include specific numbers with attribution. "According to [Source], [statistic]." AIO prefers citeable, specific claims over vague assertions.
- Publication Date: Include a visible publication date and last-updated date. AIO deprioritizes undated content for time-sensitive queries.
- Author Byline: Display author name with credentials. Link to an author page with bio, credentials, and sameAs links.
- Page Depth: Keep target pages within 3 clicks of homepage. AIO rarely cites deep, orphaned content.
Scoring Rubric (0-100)
| Criterion | Points | How to Score | |
|---|---|---|---|
| Ranks for target queries / covers fan-out sub-queries | 20 | 20 if top 10, 10 if top 100 or strong sub-query coverage, 0 if invisible | |
| Question-based headings present | 10 | 2 points per question heading, max 10 | |
| Direct answers after headings | 15 | 3 points per direct answer, max 15 | |
| Tables present for comparison data | 10 | 10 if tables used appropriately, 5 if partial, 0 if absent | |
| Lists for processes/features | 10 | 10 if present, 5 if partial | |
| FAQ section with 5+ questions | 10 | 10 if 5+, 5 if 1-4, 0 if none | |
| Statistics with citations | 10 | 2 points per cited stat, max 10 | |
| Publication/updated date visible | 5 | 5 if both dates, 3 if one, 0 if none | |
| Author byline with credentials | 5 | 5 if full byline, 3 if name only, 0 if none | |
| Clean URL + heading hierarchy | 5 | 5 if H1>H2>H3 clean, 3 if minor issues, 0 if broken |
Platform 2: ChatGPT Web Search
How ChatGPT Selects Sources
- Retrieval runs on the Bing index — 87% of ChatGPT citations match Bing results, and there is NO Google anywhere in the pipeline (Subscribe PR, Jul 2026 — ⚠️ single source, but mechanism-consistent). Bing indexation + IndexNow is the hard prerequisite for ChatGPT visibility
- Citation ref_type hierarchy (Ahrefs, 1.4M prompts, Apr 2026): search 88.46%, news 12.01%, reddit 1.93%, youtube 0.51%, academia 0.40%
- Reddit is consensus-shaping, not citation-earning. Reddit is retrieved at scale but cited at only 1.93% of ref_types, and its citation share collapsed from ~60% to ~10% in Sept 2025 (5WPR State of AI Citations, May 2026). Reddit shapes what models SAY about your brand (training/consensus layer), not what they LINK to
- ChatGPT heavily weights entity recognition — if your brand exists as a structured entity (Wikipedia, Wikidata, Crunchbase), it is far more likely to be cited
- Branded web mentions are the strongest measured off-site signal (r=0.664 — Ahrefs 75K brands, Jul 2026); raw backlink counts are the weakest measured play (⚠️ precise 0.218 figure has no locatable primary source — directional only, never quote in client/public copy)
- Freshness bias is the strongest of any platform: cited URLs average 25.7% fresher than Google organic results (Ahrefs, 17M citations, 2025-26) — NOT the viral "4.3x" figure, which is untraceable
- Title and URL slug matter before content is even read: cited-URL titles score 0.656 cosine similarity to fan-out queries vs 0.484 for non-cited; natural-language slugs are cited at 89.78% vs 81.11% (Ahrefs, Apr 2026)
Optimization Checklist
- Wikipedia Presence: Check if the brand/person/product has a Wikipedia article. If not, assess notability criteria. If notable, create a draft. If an article exists, ensure it is accurate and current.
- Wikidata Entity: Verify the entity exists on Wikidata (wikidata.org). If not, create a Wikidata item with key properties: instance of, official website, social media links, founding date, headquarters location.
- Bing Webmaster Tools: Verify the site is registered in Bing Webmaster Tools. Submit sitemap. Check for crawl errors. Enable IndexNow so new and updated pages hit the Bing index in near-real-time — this is the hard prerequisite for ChatGPT citation (87% of citations match Bing results).
- Bing Index Coverage: Use
site:domain.comon Bing to verify key pages are indexed. Bing may have different indexed pages than Google — and Google indexation alone does nothing for ChatGPT. - Reddit Consensus: Check for brand mentions on Reddit. The goal is shaping what models SAY about the brand (Reddit feeds the consensus/training layer), not earning citations — Reddit is only 1.93% of ChatGPT ref_types (Ahrefs, Apr 2026). Assess whether the brand participates authentically; paid or seeded mentions are spam-filtered and do not correlate.
- YouTube Presence: Verify YouTube channel exists with relevant content. Video descriptions should contain full URLs and entity information.
- Brand Mention Profile: Audit branded web mentions across publications, forums, and reviews — branded mentions correlate at r=0.664 with AI visibility while raw backlink counts are far weaker (⚠️ the 0.218 figure has no locatable primary source — directional only) (Ahrefs 75K brands, Jul 2026). Do not pitch press-release wires: wire pickups are ~0.04% of citations (BuzzStream, 4M citations) and are spam-filtered.
- Entity Consistency: Brand name, founding date, leadership, and key facts must be consistent across Wikipedia, Crunchbase, LinkedIn, and the official website.
- Comprehensive Content: Pages targeting ChatGPT citation should be 2000+ words with thorough topic coverage. ChatGPT prefers single authoritative sources over combining multiple thin pages.
- Clear Attribution: Include "About" sections, company descriptions, and founding stories. ChatGPT uses these for entity grounding.
Scoring Rubric (0-100)
| Criterion | Points | How to Score | |
|---|---|---|---|
| Wikipedia article exists and is accurate | 20 | 20 if exists, 10 if stub, 0 if none | |
| Wikidata entity with 5+ properties | 10 | 10 if complete, 5 if basic, 0 if none | |
| Bing index coverage of key pages | 10 | 10 if full, 5 if partial, 0 if poor | |
| Reddit consensus presence (authentic) | 10 | 10 if active discussions, 5 if mentions, 0 if none — shapes model output, not citations | |
| YouTube channel with relevant content | 10 | 10 if active, 5 if present but sparse, 0 if none | |
| Branded web mentions (publications, forums, reviews) | 15 | 3 points per mention category with authentic coverage, max 15 | |
| Entity consistency across platforms | 10 | 10 if consistent, 5 if minor discrepancies, 0 if major | |
| Content comprehensiveness (2000+ words) | 10 | 10 if thorough, 5 if adequate, 0 if thin | |
| Bing Webmaster Tools configured | 5 | 5 if verified, 0 if not |
Platform 3: Perplexity AI
How Perplexity Selects Sources
- Runs its own index plus an L3 reranker, so Bing is not the hard gate here the way it is for ChatGPT
- Historically Reddit-heavy, but Reddit's citation share collapsed from ~60% to ~10% in Sept 2025 (5WPR, May 2026) — treat community presence as consensus-shaping that influences what the model says, not as a citation farm
- Perplexity places the heaviest emphasis on community validation of all AI search platforms
- Cites multiple sources per answer (8.3 URLs on average), so there is more opportunity for mid-authority sites to appear
- Strong freshness weighting and a documented preference for page quality over domain authority — list and comparison pages dominate (May-2026 findings still hold)
Optimization Checklist
This is the start of the file. Read the full SKILL.md on GitHub.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/geo-platform-optimizer/SKILL.md
Popularity
- Repository stars
- 25
- Repository forks
- 6
- 7-day trend score
- 16
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/thesmokedev/geo-skills/geo-platform-optimizer)