How to appear in ChatGPT, Perplexity & Gemini — the complete playbook
Five practitioner methods, backed by the only controlled study of GEO signals to date. Every tactic sourced. Every claim attributed. No fluff.
Brands that implement all five methods see 2.4× higher AI citation rates than those using one or two. Princeton / KDD 2024.
AI answer engines don't cite randomly. Every citation is the output of a retrieval and preference system — and that system has measurable, reproducible signals. In 2024, researchers at Princeton ran the only controlled study to date that isolated which content changes actually move AI citation rates. The results are the foundation of this playbook.
The five methods below are listed in order of measured impact. Start with method 1. The lift compounds when you stack them.
01 — Cite your sources
Adding cited, hyperlinked sources to your content lifted AI citation rates by +41% in controlled tests. The intuition is sound: AI models are trained to trust content that demonstrates epistemic rigour. A claim with a linked source is a different signal from an unsupported assertion.
02 — Lead with statistics
Content with specific, sourced numbers earned +32% more citations than equivalent prose without figures. "AI search now drives 23× higher conversion rates than organic traffic" is infinitely more citable than "AI search converts better." The specificity is the signal.
03 — Add expert quotations
Named expert quotes produced the single largest individual lift in the Princeton study — larger than statistics and citations combined in some categories. Models are trained on human-written content where credible attribution is a trust signal.
"If you can't show the prompts, you can't show the share. Sampling is the methodology."
Name the expert, their role and affiliation. Vague attribution ("an industry expert noted...") provides no signal. Named, verifiable quotes are what models extract and cite.
04 — Structure for retrieval
Retrieval layers extract chunks of text, not full pages. If your content isn't structured for chunking, it can be indexed but not cited. Schema markup improves LLM discoverability by 67% (Yext, 2026). Clean heading hierarchy and short paragraphs enable precise extraction.
05 — Stay fresh
85% of AI citations are from content less than 2 years old. Updated content appears 4.3× more often in AI answers than stale equivalents. Freshness isn't just a nice-to-have — it's a primary preference signal, especially for Perplexity and Copilot which use real-time web search.
Establish a quarterly content review cycle. Update statistics, refresh dates, add new citations. A genuinely updated timestamp is far more valuable than artificially changing a date with no content changes.
06 — Remove access barriers
This isn't in the Princeton study but is a prerequisite: AI crawlers must be able to access your content. Check your robots.txt allows GPTBot, ClaudeBot, PerplexityBot and Google-Extended. Hard paywalls that block crawlers prevent indexing entirely.
User-agent: GPTBot Allow: / User-agent: ClaudeBot Allow: / User-agent: PerplexityBot Allow: / User-agent: Google-Extended Allow: /