AI Platform Optimization: The Complete Guide — Designing for Citations from Claude / ChatGPT / Gemini

• The "shape of information that gets cited" differs per AI. A uniform SEO-style approach makes it hard to optimize for all three AIs at once.
• From GEO Meter's observations, Claude / ChatGPT / Gemini show clearly different citation patterns:
  • Claude: Reacts strongly to llms.txt and Schema.org; prefers structured information
  • ChatGPT (via SearchGPT): FAQ / Q&A formats are easy to cite; tightly coupled to Web search results
  • Gemini: Heavily influenced by Google Search; sits on the extension of traditional SEO
• Given these differences, it is critical to separate the "shared measures" that work for all three from the "AI-specific measures" you layer on per platform.

1. The shared foundation (works for every AI)

As a baseline, there is a "shared foundation" that improves citation rates for every AI. See GEO Implementation: The Complete Guide for details.

Jumping straight to AI-specific optimization without these is doing things in the wrong order. First, ship 80% of the shared foundation, then move on to the AI-specific measures in this guide.

2. Optimization for Claude (Anthropic)

Characteristics of Claude's citation behavior

• High accuracy when reading llms.txt (in GEO Meter observations, Claude's citation rate climbs +25–30pp for sites with llms.txt)
• Interprets Schema.org structured data accurately
• Tends to "cite precisely" from the body — quoting the original passage rather than summarizing
• Strongly prefers official sources (a company's official HP / official docs)

Measures that work well for Claude

1. Raise the precision of llms.txt

Compared to other AIs, Claude reads llms.txt strictly. Avoid abstract phrasing and narrow it to 5–10 concrete URLs plus descriptions.

✅ An llms.txt example that works for Claude:

> Sencha Lab is a sencha subscription service for businesses. From ¥3,800/month, with 200 corporate customers.

## Key sections

- [Pricing plans](https://...): Two plans at ¥3,800 / ¥7,600 per month
- [Customer stories](https://...): Monthly consumption and ratings from 5 corporate customers
- [FAQ](https://...): 30 questions on delivery / cancellation / tea leaves

2. Write Article Schema accurately

Claude interprets Schema.org types such as Article, FAQPage, and Product accurately. Always include author / datePublished / dateModified.

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Five benefits of introducing sencha tea at your company",
  "author": { "@type": "Person", "name": "Taro Yamada" },
  "datePublished": "2026-04-15",
  "dateModified": "2026-05-20",
  "publisher": { "@type": "Organization", "name": "Sencha Lab" }
}

3. Publish first-party information from your official site

Claude values "the credibility of the citation source." It tends to prefer first-party information on the official HP over information funneled through third-party media.

→ Beefing up press releases, product pages, and the official blog under your own domain is highly effective.

4. Allow ClaudeBot

Allow ClaudeBot and Claude-Web in robots.txt.

User-agent: ClaudeBot
Allow: /

User-agent: Claude-Web
Allow: /

3. Optimization for ChatGPT (OpenAI)

Characteristics of ChatGPT's citation behavior

• Performs Web search via SearchGPT (the default behavior from late 2024 onward)
• Combines search results with LLM knowledge to answer → influenced by search rankings
• Strongly tends to extract and cite information in FAQ / Q&A format
• Often cites Wikipedia and third-party media (composite sources rather than the official HP alone)

Measures that work well for ChatGPT

1. Expand FAQ pages (with FAQPage Schema)

ChatGPT has a marked tendency to cite Q&A from FAQ pages verbatim. Write a single clear A per Q, in natural language.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How long can sencha tea be stored?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "We recommend consuming it within 1 year of the production date before opening, and within 1 month after opening."
      }
    }
  ]
}

2. A two-track strategy with search SEO

ChatGPT (SearchGPT) often selects citation candidates from Bing search results as the base. Because indexing and ranking on Bing have indirect effects, traditional SEO cannot be ignored.

→ Registering sitemap.xml in Bing Webmaster Tools and optimizing <title> / <meta description> remain important.

3. Aggressively earn third-party media coverage

ChatGPT tends to cite a composite of official + third-party media + Wikipedia.

→ Deliberately build up press distribution (PR TIMES and similar), contributions to note (a major Japanese long-form publishing platform), placements in industry media, and inbound citations into Wikipedia's references section.

4. Allow GPTBot / OAI-SearchBot

User-agent: GPTBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

4. Optimization for Gemini (Google)

Characteristics of Gemini's citation behavior

• Deeply integrated with Google Search (an extension of traditional SEO)
• Schema.org and the Knowledge Graph have an outsized impact
• Citation URLs are handed off through Google's vertexaisearch redirect
• Tends to cite Google Business Profile and local search information (especially strong in regional / local-business categories)

Measures that work well for Gemini

1. Comprehensive Schema.org implementation

Gemini feeds Schema.org into the Google Knowledge Graph, so structured data has the most direct impact.

Priority order:

1. Organization Schema (company info)
2. Product / Service Schema (the products / services you offer)
3. FAQPage Schema
4. Article Schema (article pages)
5. LocalBusiness Schema (stores / regional businesses)

2. Maintain your Google Business Profile

For local industries (tax accountants, beauty salons, restaurants, etc.), maintaining a Google Business Profile directly drives Gemini citations.

• Register business hours / address / phone (NAP info) on Google Business
• Acquire reviews
• Implement LocalBusiness Schema on the official HP and link to the Google Business URL

3. Stay rigorous with traditional SEO

Because Gemini operates as the evolved form of Google Search, traditional SEO measures apply directly.

• Optimize page titles and meta descriptions
• Build a sound internal linking structure
• Mobile responsiveness / Core Web Vitals
• Earn inbound links

4. Allow Google-Extended

User-agent: Google-Extended
Allow: /

Google-Extended is the UA Gemini uses to access content for training and citation. Disallowing it removes you from Gemini's citation pool.

5. AI-specific priority summary

6. Why per-AI observation matters

Cases like "cited on Claude but not on ChatGPT" come up routinely in GEO Meter data. If you only look at one AI, you risk pointing your strategy in the wrong direction.

GEO Meter observes Claude / ChatGPT / Gemini in parallel and lets you compare citation counts, rankings, and context per AI.

→ Track all three AIs monthly with GEO Meter

7. Common misconceptions

❌ "If Claude cites you, every AI will cite you"

Each AI's citation logic differs, so being #1 on Claude but invisible on ChatGPT happens regularly. You need to address all three separately.

❌ "SEO alone is enough for AI too"

For Gemini, somewhat true. But Claude / ChatGPT are less influenced by SEO. The llms.txt / Schema.org / FAQ trio comes first.

❌ "You don't need to optimize for Perplexity"

Perplexity is not currently in GEO Meter's observation scope, but measures aimed at Claude / ChatGPT (llms.txt + Schema.org + third-party media) carry over almost directly. Treat it as medium priority.

8. Summary

• Citation behavior differs per AI, but the shared foundation (llms.txt + Schema.org + FAQ) works for every AI
• Claude → llms.txt precision + first-party publishing + Article Schema
• ChatGPT → FAQ expansion + third-party media coverage + traditional SEO
• Gemini → Comprehensive Schema + Google Business + traditional SEO
• Without per-AI observation across the three, your strategy will misfire → monthly measurement is non-negotiable

Related resources

• GEO Basics: The Complete Guide — Big picture and positioning of GEO
• GEO Implementation: The Complete Guide — Priority order and implementation of the five measures
• llms.txt Implementation: The Complete Guide — How to write llms.txt
• Free GEO Meter Diagnostic — Measure your current state across all three AIs