Search/retrieval AI crawlers can reach your site

The #1 reason for zero AI citations is that crawlers can't access the page. OAI-SearchBot, Claude-SearchBot and PerplexityBot must not be blocked.

robots.txt AI-bot directives are deliberate, not accidental

Many sites silently block GPTBot/ClaudeBot in boilerplate robots.txt; decide separately for training bots vs search/retrieval bots.

You separate training bots from search/retrieval bots

OpenAI (GPTBot vs OAI-SearchBot/ChatGPT-User) and Anthropic (ClaudeBot vs Claude-SearchBot/Claude-User) use distinct user agents — block training without blocking citation crawlers.

Google-Extended decision is made consciously

Google-Extended controls Gemini/Vertex training but does NOT affect AI Overviews eligibility, which follows normal Googlebot indexing.

Pages are indexed and eligible for Google AI Overviews

97% of AI Overview citations come from the top 20 organic results — standard indexability is the entry ticket.

Primary content renders in HTML without client-side JS

Most AI crawlers do not execute JavaScript; content hidden behind hydration or tabs is invisible to them.

An llms.txt file points AI systems to your key content

A Markdown index of your core pages helps AI systems navigate and may be retrieved during web-enabled RAG answers.

An ai.txt declares your AI usage and access preferences

An explicit policy file complements robots.txt and signals intentional, machine-readable terms to AI agents.

Content is dense with concrete statistics and numbers

The Princeton GEO study found adding statistics raised position-adjusted visibility by up to ~40%, strongest in Law, Government, Debate and Opinion topics.

You include relevant expert quotations

Quotation Addition was the single best-performing GEO method (~41% on position-adjusted word count; +22% on live Perplexity tests).

Claims are backed by visible, linked source citations

Citing sources delivered 30-40% visibility gains in the GEO study and is strongest for factual queries.

Writing uses confident, authoritative language

Authoritative phrasing improved citation visibility, especially in Debate, History and Science domains.

Prose is fluent and clearly written

Fluency optimization gave 15-30% gains — generative engines value presentation, not just facts.

You avoid keyword stuffing

Keyword stuffing gave little-to-no gain and tested ~10% worse than baseline on Perplexity — old SEO tricks hurt GEO.

Pages contain quotable, self-contained statements

AI engines lift discrete sentences; standalone, declarative claims are far more citable than context-dependent ones.

Each key page has at least one specific, attributed data point

Adding one named-source statistic per page is a top recommended move for earning citations.

Content splits into clean, self-contained passages

AI retrieval works on chunks; sections that stand alone become independently citable units — restructuring can yield 2-4x more citations.

Key information appears in the first ~300 words

Front-loading the answer is a top recommended fix; AI extractors favour the lead of the page.

Clear H2/H3 hierarchy bounds each topic

AI systems chunk content along heading and list boundaries; clean hierarchy improves ChatGPT extraction.

Headings mirror the follow-up questions users ask

Matching H2/H3s to natural questions aligns chunks with how AI engines query your content.

Each passage reads correctly when pulled out of context

Citations show fragments; passages must define their subject (avoid 'it', 'this') to be safely quoted.

Definitions are stated plainly ('X is …')

Explicit definition blocks are easy for engines to extract as authoritative answers.

Comparisons use clean, parseable tables

Structured tables let AI engines extract and reproduce comparison data directly.

Ranked lists / listicles cover comparison & 'best' queries

Listicle-format pages are heavily favoured for 'best X' and recommendation queries in AI answers.

Your page converts cleanly to Markdown

Markdown is the native format LLMs ingest; messy HTML degrades chunk quality and citability.

Pages show clear Experience, Expertise, Authority & Trust

AI engines weight source trust heavily before citing; weak E-E-A-T pages are filtered out.

Content has named authors with credentials and bios

Identifiable, qualified authors raise trust signals AI systems use to choose sources.

Your claims are corroborated across independent sources

AI recommendation engines build multi-source consensus — they cite brands that multiple trusted sites agree on.

You earn links/mentions from high-authority domains

Perplexity especially favours data-rich content from high-authority sources; off-site authority drives citations.

You have authentic presence on Reddit and forums

Reddit is among the most-cited sources across major AI platforms; genuine contributions earn high-trust citations.

Key topics are covered with supporting video content

YouTube overtook Reddit as the most-cited social platform in AI answers in early 2026.

You publish original research, data or surveys

Unique, quotable statistics make you the primary source AI engines must cite rather than a paraphraser.

Your brand exists as a Wikidata entity

Wikidata is a primary input to Google's Knowledge Graph and has no notability threshold — the most powerful sameAs target.

Your brand is eligible for a Knowledge Graph panel

A recognised entity is far more likely to be named accurately in AI answers.

Organization schema declares your brand as an entity

Machine-readable Organization markup tells AI who you are, what you do and where you exist online.

sameAs links your entity to Wikidata, Wikipedia & profiles

sameAs is the single most important entity signal — it lets AI triangulate your identity across trusted sources.

Brand name and details are consistent across the web

Inconsistent naming fragments your entity and reduces confident AI attribution.

You cover the entities and subtopics your niche expects

Topical entity completeness signals depth; gaps let competitors own the citation.

knowsAbout / expertise is declared in schema

knowsAbout disambiguation strengthens Knowledge Graph recognition and topical authority.

Structured data is present and valid

AI engines use schema to verify entity facts and increase accurate citations.

Schema coverage is complete for each page type

Stacked, relevant schema (Article, Product, FAQ, HowTo) gives engines machine-readable answer blocks.

Article schema includes datePublished and dateModified

Explicit dates let AI engines assess freshness, which strongly influences citation choice.

Author/Person schema connects content to real experts

Person markup ties E-E-A-T signals to a verifiable entity AI can trust.

Content shows recent, accurate published/updated dates

AI engines cite content ~25% fresher than classic search; pages updated within 2 months earn ~28% more citations.

Priority pages are refreshed on a regular cadence

Perplexity in particular favours recently updated content; a 7-14 day refresh cycle is recommended for competitive topics.

Outdated facts and stats are corrected or removed

Stale or wrong figures undermine the trust signals AI relies on to cite you.

Overall GEO score is measured and improving over time

A baseline score lets you prove which changes actually move AI visibility.

LLM-readiness (LLMO) is scored for key pages

Pages must be machine-readable and extractable to be used by LLM answers.

You test whether AI assistants name your brand

A mention (named) is distinct from a citation (linked); track both to understand visibility.

You track citations and mentions in ChatGPT

ChatGPT leans on training data and only searches on demand — monitor how it represents your brand.

You track citations in Perplexity

Perplexity always cites linked sources and runs a live search per query — a fast feedback loop for GEO.

You track visibility in Gemini / Google AI

Gemini rewards strong Google organic rankings; track whether AI Overviews and Gemini surface you.

You monitor share of voice vs competitors in AI answers

Share-of-voice across ChatGPT, Perplexity, Gemini and Copilot shows where rivals out-cite you.

Snippets are tested against AI-mode answer formats

Validating how content renders as an AI answer catches extraction problems before they cost citations.