How to Optimize for Claude and Perplexity (Not Just ChatGPT)

Most GEO advice focuses on ChatGPT. That's a mistake.

ChatGPT dominates AI traffic today (78% market share), but Perplexity is growing at 15% and Claude users spend 19 minutes per session—nearly 4x the organic search average.

I learned this building Loamly: optimizing for one AI platform doesn't mean you're optimized for all of them. They work differently, cite differently, and value different content characteristics.

---

What the Loamly Dataset Shows

From 1,528 company reports (Jan 9-21, 2026):

Platform	Avg Citation Rate
ChatGPT	0.703
Claude	0.643
Gemini	0.498

Brand authority matters more on Claude than anywhere else:

Platform	Brand Authority Correlation	GEO Correlation
ChatGPT	0.320	0.114
Claude	0.430	0.057
Gemini	0.180	0.039

This is the core optimization point: Claude is authority-sensitive, Gemini shows weak correlations across the board, and ChatGPT is mostly training-data-driven. Treat them as separate systems.

Why Each Platform Matters

Platform	Traffic Share	Session Duration	Citation Style
ChatGPT	77.97%	10 min	Inconsistent, training-biased
Perplexity	15.10%	Variable	Real-time, heavily cited
Claude	0.17%	19 min	Depth-focused, context-rich
Gemini	6.40%	Variable	YouTube/multi-modal preference

Source: SE Ranking AI Traffic Study

The conversion story is more important than volume. AI visitors convert at 11x the rate of organic search. When Claude sends a visitor who stays 19 minutes, that's not a casual browser—that's a qualified lead.

Understanding the Architectural Differences

This is the key insight most marketers miss: these platforms don't work the same way.

ChatGPT: Training Data + Optional Search

ChatGPT primarily generates answers from patterns learned during training on massive datasets. It mentions brands roughly 3x more often than it formally cites them because many responses come from training data without web retrieval.

When ChatGPT does use web search (via "Browse with Bing"), it becomes more like Perplexity. But most ChatGPT interactions don't trigger web retrieval.

Implication: Being in ChatGPT's training data matters more than real-time optimization.

Perplexity: Real-Time Search First

Perplexity is built entirely around retrieval-augmented generation (RAG). Every answer goes through live web search first. The platform searches the internet, gathers information, then synthesizes answers with numbered citations.

This means:

• Your content's freshness directly impacts citation likelihood
• Being indexed by Perplexity matters more than training data
• Citations are consistent and traceable

Implication: Real-time optimization works. Update your content regularly.

Claude: Extended Context + Depth

Claude's architectural advantage is its 200,000-token context window—roughly 150,000 words. Claude can process entire research papers, comprehensive guides, and multi-document analyses without chunking.

Claude values:

• Comprehensive, thoroughly researched content
• Sophisticated analysis and expert reasoning
• Content that demonstrates nuance and depth

Implication: Depth wins. Create the most comprehensive resource on your topic.

Platform-Specific Optimization Strategies

Optimizing for Perplexity

Perplexity's real-time web retrieval creates specific optimization opportunities.

1. Content Freshness Is Critical

According to research on Perplexity's citation patterns, content decay begins within 2-3 days of publication. This is dramatically different from traditional SEO, where monthly or quarterly updates suffice.

For Perplexity optimization:

• Audit high-performing pages monthly
• Add new data, insights, or sections regularly
• Republish with updated timestamps
• Prioritize pages showing declining visibility

2. The Reddit Phenomenon

This surprised me when I first saw the data: Reddit accounts for 46.7% of Perplexity's citations. That's nearly half.

Why? Perplexity values:

• Real-time community discussions
• Authentic, unfiltered user experiences
• Practical implementation details from actual users
• Peer recommendations over institutional authority

Action items:

• Participate authentically in relevant subreddits
• Monitor Reddit for brand discussions
• Extract Reddit insights into your own content (with attribution)

3. Answer-First Formatting

Pages with FAQ schema appear in AI responses at 3x the rate of equivalent unstructured content.

Structure content with:

• Direct answers in opening paragraphs
• Clear question-answer pairs
• Comparison tables with original data
• Numbered steps for processes

4. Conversational Query Optimization

Perplexity encourages users to ask questions naturally. Optimize for conversational queries, not keyword phrases:

Instead of: "best email marketing software" Optimize for: "What's the best email marketing software for a small B2B SaaS company?"

Optimizing for Claude

Claude's approach requires different tactics.

1. Go Deeper Than Competitors

Claude's extended context means it can process and prefer unusually comprehensive content. Where traditional SEO might reward concise featured-snippet-friendly content, Claude rewards depth and nuance.

Create:

• Guides exceeding typical word counts (3,000-5,000+ words)
• Research with extensive supporting data
• Documentation covering edge cases and variations
• Content demonstrating sophisticated reasoning

2. Author Credentials Matter More

Claude appears to weight author authority heavily. According to research on author bylines, AI systems preferentially cite sources with:

• Detailed author biographies
• Verifiable credentials (LinkedIn, ORCID IDs)
• Professional directory listings
• Transparent editorial review processes

Action: Create dedicated author profile pages with credentials, not just generic "staff writer" bylines.

3. Control Your Crawler Access

Anthropic operates three distinct crawlers with different purposes:

Bot	Purpose	Respects robots.txt
ClaudeBot	Training data collection	Yes
Claude-User	Real-time user queries	No*
Claude-SearchBot	Search result quality	Yes

*Claude-User ignores robots.txt because users initiated the request.

Strategic choice: Block ClaudeBot (prevent training data collection) while allowing Claude-User (enable live search access) if you want to control training but maintain visibility.

User-agent: ClaudeBot
Disallow: /

User-agent: Claude-User
Allow: /

Optimizing for Both Simultaneously

The good news: many optimizations benefit multiple platforms.

Schema Markup (Universal)

Implement structured data that helps all AI systems:

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "How do I track AI traffic to my website?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Use a tool like Loamly that specifically detects AI referrals, or configure GA4 with custom regex matching for known AI platform domains."
    }
  }]
}

Essential schema types:

• FAQ schema: Question-answer pairs (highest impact)
• Article schema: Publication dates, authors, modification dates
• Organization schema: Brand entity definition
• Person schema: Author credentials and expertise

Semantic HTML (Universal)

Both Claude and Perplexity benefit from clean document structure:

<article>
  <h1>Main Topic</h1>
  <section>
    <h2>Subtopic</h2>
    <p>Content...</p>
  </section>
</article>

Use proper semantic tags:

• <h1> through <h4> in logical hierarchy
• <article>, <section>, <aside> for structure
• <ul> and <ol> for lists
• <table> for comparison data

Original Data (Universal)

Content with 19+ data points receives 5.4 average citations vs 2.8 for minimal data content.

Create:

• Original research and surveys
• Benchmark reports with primary data
• Case studies with specific metrics
• Comparison tables with tested results

What Actually Differs by Platform

Factor	Perplexity Priority	Claude Priority
Freshness	Critical (2-3 day decay)	Moderate
Depth	Important	Critical
Reddit presence	Very high	Moderate
Author credentials	Important	Very high
Real-time updates	Critical	Moderate
Comprehensive coverage	Important	Critical
FAQ structure	High	Moderate

The Citation Pattern Differences

Understanding how each platform cites helps you structure content appropriately.

Perplexity Citations

Perplexity provides numbered citations linked directly to sources. Users see exactly where information came from and can click through.

This means:

• Citation prominence matters (earlier = more clicks)
• Your source appears in a traceable way
• Getting cited = measurable traffic potential

Claude Citations

Claude's citation behavior is less consistent. It may:

• Reference sources conversationally without explicit links
• Cite from training data without providing URLs
• Provide links when web search is enabled

This means:

• Training data presence matters
• Being a "go-to" authority in your space compounds
• Depth and comprehensiveness build long-term citation likelihood

Measuring Cross-Platform Performance

You can't optimize what you can't measure. Here's how I track performance across platforms:

Manual Prompt Testing

Ask each platform standardized questions about your industry weekly:

• "What's the best [your category] tool?"
• "How do I [problem you solve]?"
• "Compare [your brand] vs [competitor]"

Document:

• Are you cited?
• What position in the citation list?
• What competitors appear?
• What's the sentiment?

GA4 Referral Tracking

Configure GA4 to track known AI referrers:

perplexity\.ai|claude\.ai|chat\.openai\.com|gemini\.google\.com

Create custom channel groups matching these patterns.

Caveat: This only captures clicks, not citations. You'll miss copy-paste traffic (the majority) and mentions that don't result in visits.

AI Visibility Monitoring Tools

Specialized tools track citation frequency across platforms:

• SE Ranking Visible
• BrightEdge Prism
• Semrush AI features

These measure visibility even when users don't click through.

My Honest Take

When I started building Loamly, I thought "AI optimization" was one thing. It's not.

Perplexity rewards freshness and real-time relevance. Claude rewards depth and authority. ChatGPT sits somewhere in between, biased by its training data.

The companies winning at GEO aren't optimizing for "AI" generically—they're building platform-specific strategies while maintaining content quality that works everywhere.

Start with:

1. FAQ schema on your most important pages
2. Author credentials and bylines on all content
3. Update cadence appropriate to your content type
4. Original data and research where possible

Then add platform-specific tactics based on where your audience is.

---

See Your AI Visibility Across Platforms

Not sure where you're appearing—and where you're invisible?

Start with a free Loamly AI visibility check to see how you show up across ChatGPT, Perplexity, Claude, and Gemini.

Different platforms, different opportunities.