What it means
An LLM citation is the inline source attribution an AI engine shows beside or beneath a generated answer. It points to the chunks the retriever pulled and the generator chose to credit, each with a clickable URL. Not every source becomes a citation: A model can only cite pages that make it into the answer's retrieval set, and knowledge absorbed during pre-training usually goes unattributed. Live retrieval is what adds the URL that makes attribution possible.
Why it matters
Higher rankings raise your citation odds, but many AI answers cite nothing at all, so visibility no longer guarantees a click. Because many LLMs use search engines as a retrieval source, ranking in Google's top 10 raises your probability of entering the candidate pool and getting cited. Yet citation rates stay low even when models reach relevant sources: One study Kevin cites found Gemini gives no clickable citation in 92% of answers, and about a quarter of ChatGPT responses fetch no live content at all.
Say a mid-market B2B SaaS company sits at position 8 for its priority queries and gets cited in roughly 1 of 10 tracked AI answers. It rewrites those pages to rank in the top 3 across a wider set of question variations, and its citation rate doubles to 2 of 10.
How to use this knowledge
Rank in the top 10 for the question and long-tail variations around your core topics, since those pages have the highest citation probability. Track citations as a distinct metric from rankings and traffic, because you can be cited without earning a click. Watch model-level behavior: Routers that favor non-reasoning models show fewer citations and send less traffic, so referral swings are not always your fault. Make your pages easy to retrieve and quote so they survive into the answer the generator builds.
Growth Memo guidance
Higher organic rankings increase the probability of entering the LLM's candidate pool and receiving citations. — State of AI Search Optimization 2026
The router favors non-reasoning models, which show fewer citations and send less traffic out. — LLM traffic is shrinking
Grounding varies from model to model and not all LLMs prioritize pages ranking at the top of Google search. — How much can we influence AI responses?
AI mention — the unlinked cousin of a citation, where your brand is named without a clickable URL.
Candidate pool — the set of pages eligible for citation, which top-10 rankings help you enter.
Grounding — the retrieval step that pulls live pages into an answer and makes attribution possible.
Query fan-out — the expansion of one prompt into many retrievals, widening which pages can be cited.
Zero-click — the outcome when an answer resolves without any citation click.