What are the actual ranking factors for AI search? Not Google - ChatGPT, Perplexity, etc.

I’ve spent 18 months researching this. Here’s what we know:

Primary AI Citation Factors:

What matters LESS than in Google SEO:

• Backlink quantity (though quality still matters)
• Keyword density/optimization
• Page speed (still matters, but less weighted)
• Internal linking structure

What matters MORE than in Google SEO:

• Brand/entity recognition
• Author expertise signals
• Content extractability
• Cross-platform consistency

The key difference: Google ranks pages. AI cites passages. You need content that has citation-worthy passages.

Let me break down how AI systems seem to evaluate authority:

Author-level signals:

• Named author with bio
• Credentials relevant to topic
• Track record (other published work)
• Social proof (followers, recognition)

Brand-level signals:

• Brand recognition (Wikidata/Knowledge Graph)
• Consistent NAP across platforms
• Press coverage and mentions
• Social presence and engagement

Content-level signals:

• Citations to authoritative sources
• Original research/data
• Expert quotes or contributions
• Comprehensive, accurate information

Cross-web signals:

• Mentioned by other authoritative sites
• Cited in forums (Reddit especially)
• Referenced in news/publications
• Consistent information across sources

What this means practically:

This is why Reddit gets cited despite low “traditional” authority - it has strong cross-web mention patterns and authentic user signals.

Each AI platform weights factors differently. Here’s what I’ve observed:

ChatGPT:

• Heavy weight on training data (historical authority)
• Brand recognition matters a lot
• Prefers comprehensive, established sources
• Real-time search follows Bing signals

Perplexity:

• Real-time relevance is paramount
• Recency weighted heavily
• Source diversity matters (cites multiple sources)
• Strong Reddit/forum weighting

Google AI Overview:

• Google rankings are primary signal
• E-E-A-T signals from Google’s evaluation
• Structured data helps
• Follows traditional SEO patterns most

Claude:

• Training data only (no real-time search yet)
• Authority from training corpus
• Hard to optimize directly

The implication: You can’t optimize for “AI” generically. Each platform is different.

The unifying factors: Content quality, clarity, and expertise help everywhere. Start there.

On structured data specifically:

Does schema markup help AI visibility?

Our testing (200 pages):

Why structured data helps:

1. Helps AI parse content structure
2. Clarifies entity relationships
3. Signals content type and purpose
4. Makes extraction more reliable

Most impactful schema: FAQPage and HowTo - these directly align with question-answer format AI prefers.

Implementation priority:

1. FAQPage on FAQ content (biggest lift)
2. HowTo on tutorial content
3. Article schema with author info
4. Product schema for e-commerce

Caveat: Schema is a signal, not a guarantee. Great content without schema beats bad content with schema.

On the recency question:

How much does freshness matter for AI?

Depends heavily on query type:

Platform differences:

• Perplexity: Real-time, heavily weights recency
• ChatGPT Search: Current when search enabled
• ChatGPT base: Training cutoff limits recency
• Google AI Overview: Current via index

What we do:

• Update key content monthly with fresh stats
• Add “last updated” dates prominently
• Refresh seasonal content before seasons
• Keep evergreen content evergreen

The balance: For evergreen topics, authority beats recency. For current topics, recency can beat authority.

Know which category your content is in.

You mentioned Reddit’s high citation rate despite low “traditional authority.” Here’s why:

Why Reddit gets cited so much:

1. Authentic user signals

   • Upvotes = community validation
   • Comments = engagement depth
   • Questions/answers format matches AI queries

2. Training data presence

   • Heavy in ChatGPT training
   • Used by most LLM training

3. Real experience

   • Users sharing actual experiences
   • Less marketing polish, more authenticity
   • E-E-A-T “Experience” signal

4. Question-answer format

   • Already structured as Q&A
   • Easy for AI to extract
   • Matches user query patterns

What this tells us about AI “authority”:

AI systems value DIFFERENT authority signals than Google:

• Authenticity over polish
• Experience over credentials (sometimes)
• Community validation over editorial review
• Natural language over SEO language

The lesson: Traditional authority metrics (DA, backlinks) don’t tell the full story. AI systems have their own authority evaluation.

One factor that’s underrated: Entity recognition.

What is entity recognition for AI? AI systems need to understand what/who you are as an entity - a brand, organization, or author.

How AI recognizes entities:

• Wikipedia/Wikidata presence
• Google Knowledge Panel
• Consistent mentions across web
• LinkedIn, Crunchbase profiles
• Press coverage with consistent info

Why entity recognition matters: If AI systems “know” who you are, they trust you more.

Testing this: Ask ChatGPT: “What is [your brand]?”

If it gives accurate info: You have entity recognition. If it hallucinates or doesn’t know: You have a problem.

Building entity recognition:

1. Claim/create Wikidata entry (if notable enough)
2. Ensure consistent NAP across platforms
3. Get mentioned in contexts AI would learn from
4. Build Knowledge Panel through PR/Wikipedia

The ROI: Brands with strong entity recognition get cited 2-3x more than unknown brands, even with similar content quality.

Important caveat on all these “factors”:

We’re mostly observing correlations, not confirmed causation.

What we actually know:

• Certain content characteristics correlate with higher citation rates
• We can’t see inside AI systems’ actual algorithms
• Correlation ≠ causation
• AI systems change constantly

What this means: All these “ranking factors” are educated guesses based on observation.

The safe bets (correlate strongly):

• Content quality and clarity
• Query relevance
• Source authority
• Structured format

The uncertain areas:

• Exact weighting of factors
• Platform-specific algorithms
• How factors interact
• How algorithms change over time

Practical approach: Optimize for the fundamentals (quality, clarity, authority). Track results. Iterate based on what you observe.

Don’t over-optimize for “AI ranking factors” you’re not sure exist. Focus on creating genuinely valuable content.

This is comprehensive. My takeaways:

The Framework I’ll Use:

High-confidence factors to optimize:

1. Content clarity and extractability
2. Direct relevance to queries
3. Author/brand authority signals
4. Comprehensive topic coverage
5. Structured data (FAQPage, HowTo)

Medium-confidence factors:

1. Recency (topic-dependent)
2. Cross-web mentions
3. Entity recognition
4. Schema markup

Low-confidence/indirect factors:

1. Backlinks (indirect via Google rankings)
2. Traditional DA metrics
3. Technical SEO (still matters, just less)

Key differences from Google SEO:

• Less emphasis on links
• More emphasis on entity/brand recognition
• Content extractability matters more
• Platform-specific optimization matters

What I’m doing differently:

1. Adding author bios and credentials
2. Implementing FAQPage schema everywhere
3. Building entity recognition (Wikidata, mentions)
4. Tracking with Am I Cited to see what actually works

The meta-insight: AI “ranking factors” are evolving. Track, test, iterate. Don’t assume today’s factors are permanent.

Thanks everyone!