GEO/AEO Optimization 2026: What Does AI Cite?
How do AI search engines (ChatGPT, Gemini, Perplexity) select sources? Authority, structure, and citability — not PageRank. 15 patterns from 52 sources.
TL;DR
AI search engines select sources based on authority, structure, and citability — not PageRank. Schema.org structured data, FAQ sections, TL;DR blocks, and evidence-based content significantly increase AI citation probability. Traditional SEO techniques (link building, keyword density) are less effective.
Executive Brief
We examined the source selection logic of AI search engines (ChatGPT, Gemini, Perplexity, Google AI Overviews) based on 52 sources. The research question: how should we select and optimize our content so that AI systems cite it?
Key Patterns
What works:
- Schema.org markup — structured data is the “language” of AI search; FAQ, HowTo, Article, Person schemas
- TL;DR and summary blocks — AI quotes literally; if a ready summary exists, it uses that
- FAQ sections — question-answer format is ideal for AI search engines
- Authority signals — author profile, institutional backing, citations to other authoritative sources
- Evidence-based content — claims backed by sources; AI gives preference to these
What doesn’t work:
- Keyword density optimization (AI understands semantically, doesn’t search for keywords)
- Traditional link building (AI doesn’t look at PageRank)
- Clickbait titles (AI evaluates informational value, not click potential)
- Thin content across many pages (AI values depth, not quantity)
Cross-LLM differences:
- ChatGPT, Gemini, and Perplexity use different source selection logic
- Perplexity is the most transparent (explicit source attribution)
- AI Overviews work embedded within Google’s search context
- What works across all: structure, authority, evidence
Methodology
- Sources: 52 (web: 34, academic: 11, industry reports: 7)
- Research rounds: 5 (base + 3 deep dives + blind spot audit)
- Patterns: 15 identified, 11 supported, 3 contested, 1 candidate
- Blind spot audit: examined non-English AI search source selection and video content citability
Full Research
The full field report is available through consultation. Learn more about the Gestalt Research Engine →
Gestalt Research Engine
This research was produced with the GRE pipeline: systematic source collection, pattern recognition (figure/background/noise), blind spot audit, convergence check. Learn more →