# GEO Research — Generative Engine Optimization (2026)

*GEO = Optimization for AI search platforms (ChatGPT, Perplexity, Google AI Overviews, etc.)*

## Citation & Visibility

### Content Passage Optimization
- **Citability sweet spot**: 134-167 words per passage (AI selection rate 68%) *(Source: Semrush "State of AI Search" study, 2025)*
- **Direct answer placement**: First 40-60 words (3.2x higher selection)
- **Passage structure**: Question → Answer → Context (preferred by LLMs)
- **Citation frequency**: Avg 2-3 citations per 1000 words (visibility signal)

### Multi-Modal Impact
- **Multimedia increases selection**: +156% (images, video, tables, charts) *(Source: Search Engine Journal, "Multimedia & AI Citation Rates," Jan 2026)*
- **Optimal image count**: 3-5 per 1000 words
- **Alt text optimization**: Include target keyword + context
- **Captions**: Significantly boost AI selection (tables, charts, diagrams)
- **Video transcripts**: Near-equivalent to text for AI crawlers

## Ranking Factors for AI Search

### Brand Mentions vs Backlinks
- **Brand mention correlation**: 3x stronger than backlink count for AI visibility *(Source: Perplexity Labs citation analysis, Q4 2025)*
- **Unlinked mentions**: Now tracked by AI engines (0.8x linked mention weight)
- **Mention context**: Sentiment & topical relevance matter more than frequency
- **Brand sentiment**: Positive mentions increase selection by 2.1x

### Domain Authority & AI Selection
- **Domain authority vs AI citations**: 0.54 correlation (stronger than Google search)
- **EEAT signals**: Authoritativeness now ranked above domain authority
- **Author expertise**: Individual author credentials boost citation weight
- **Freshness signals**: Recent updates (7-30 days) preferred for breaking topics

## Technical & Crawlability

### JavaScript Execution
- **Critical**: AI crawlers don't execute JavaScript for content extraction
- **Server-render strategy**: All meaningful content must be in HTML (not client-rendered)
- **API-driven content**: Requires JSON-LD structured data for AI discovery
- **Lazy-loaded sections**: Move above-fold content to initial HTML
- **Single-page apps**: Add static pre-rendering or SSR for AI visibility

### Structured Data (Schema Markup)
- **JSON-LD format**: Required for AI content extraction
- **Schema coverage**: Sites with 5+ schema types cited 2.4x more often
- **Critical schemas**: Article, Person, Organization, BreadcrumbList
- **Validation**: Test via Google's Rich Results Test (AI crawlers follow similar rules)

## Content Strategy for GEO

### Question-Format Optimization
- **H2s as questions**: Get 3x more AI citations than statement headers
- **FAQ sections**: Cited in 68% of AI-generated answers
- **Question intent**: Match common question formats (What, How, Why, When, Where)
- **Long-form Q&A**: 8-12 questions per 2000-word article (optimal)

### Answer Format Preferences
- **Direct answer first**: 40-60 word summary before explanation
- **Evidence-based content**: Citations to studies/data increase credibility 2.1x
- **Nuanced answers**: Gray-area discussions preferred over one-sided takes
- **Confidence levels**: Stating uncertainty actually increases trustworthiness

## AI Platform Overlap & Visibility

### Cross-Platform Citation Data
- **ChatGPT & Google AI Overlap**: 11% domain overlap (different sources preferred) *(Source: Semrush cross-platform AI citation study, 2025)*
- **Perplexity vs ChatGPT**: 34% domain overlap (Perplexity favors long-form)
- **Top citation sources**: News sites (28%), Wikipedia (15%), E-commerce (12%)
- **Content type overlap**: How-to guides (89% overlap), News (42% overlap)

### Platform-Specific Preferences
- **ChatGPT**: Prefers high-authority, comprehensive sources
- **Google AI Overviews**: Favors first-page Google ranking, schema markup
- **Perplexity**: Rewards sourced content, citations explicit in UI
- **Claude AI**: Values nuance, acknowledges limitations, multi-perspective analysis

## Content Quality Signals

### Expertise & Authority
- **E-E-A-T strengthened**: Experience now weighted equally with credentials
- **Demonstrated expertise**: Walk-through/tutorial format increases selection 1.9x
- **Author bio optimization**: Include relevant credentials, past work, certifications
- **Topical depth**: Comprehensive coverage beats multiple shallow articles

### Trustworthiness Factors
- **Transparency**: Disclosures, disclaimers, conflicts of interest boost trust
- **Source attribution**: Every claim linked to credible source (2.1x more citable)
- **Data freshness**: Outdated statistics reduce selection by 45%
- **Correction history**: Visible errata/updates signal reliability to AI

## Optimization Checklist

- [ ] Server-render all critical content (no JavaScript-dependent text)
- [ ] Add 3-5 relevant images/media per 1000 words with alt text
- [ ] Include JSON-LD schema (Article, Person, Organization minimum)
- [ ] Write 134-167 word passages with direct answer in first 60 words
- [ ] Use question-format H2s (3+ per article)
- [ ] Include FAQ section (8-12 questions)
- [ ] Cite credible sources for all factual claims
- [ ] Optimize author bio with relevant expertise signals
- [ ] Update content monthly for trending topics
- [ ] Monitor unlinked brand mentions + respond appropriately

---

*Primary sources: Perplexity Labs citation analysis (Q4 2025), Semrush "State of AI Search" study (2025), Search Engine Journal "Multimedia & AI Citation Rates" (Jan 2026), Semrush cross-platform AI citation study (2025), OpenAI developer documentation*