What Is an AI Citation?
What is an AI citation?
An AI citation is when an assistant points to a specific source behind its answer: a link, a footnote, an "according to...". Perplexity does this out in the open; ChatGPT and Gemini do it when they browse the live web. A citation is the model showing its work: this is the page it leaned on to say what it said.
For a brand, citations are gold, because they're a map. If a competitor is cited for "best CRM for startups" and you aren't, the citation tells you exactly which page earned that mention, and which page you'd need to be on.
AI citations vs. mentions
Worth separating these. A mention is your brand name appearing in an answer. An AI citation is a source the model links to. You can be mentioned without being cited (the model names you from memory) and a page can be cited without your brand being the star of it, like a roundup that got cited. The pairing is what matters: a mention backed by a citation to your own page is the strongest signal you can get.
Why AI citations matter
Because they're the closest thing to a paper trail in a black box. Most of what drives an AI answer is invisible. Citations are the visible part: the handful of sources the model was willing to name. Track which domains keep getting cited in your category and you've found the pages that actually shape the answer. That's where the work is.
How to analyze your AI citations
Collect the answers for the questions you care about, pull the cited sources, and look for the pattern: which domains recur, whether you're among them, and which competitor pages keep showing up. We walk through the full method in AI citation analysis. If you'd rather have citations tracked automatically alongside your other signals, that's part of what our scoring measures.
Which engines cite, and how
Citation behavior differs sharply by engine, and the differences shape where citation work pays off:
| Engine | Citation behavior |
|---|---|
| Perplexity | Cites almost every answer; favors fresh, linkable pages |
| Google AI Overviews | Shows source links; leans on indexed, ranking pages |
| Gemini | Cites with grounding on; pool close to Google's index |
| ChatGPT | Cites only in sourced/search mode; reads a narrower pool |
| Claude, Grok | Cite when web search is active; otherwise answer from training |
The practical consequence: a citation strategy tuned for Perplexity (fresh comparison content) differs from one tuned for AI Overviews (classic ranking strength). Measure per engine before deciding where to invest.
The SEO parallel that mostly holds
Citations are to AEO what backlinks are to SEO (the currency of being trusted as a source), with one important twist: engines cite pages that answer, not pages that are merely authoritative. A DR-90 homepage gets cited less than a DR-40 comparison article that directly answers the question. That inverts some SEO instincts: for citations, the answering page beats the important page.
Quick wins that earn citations
Three moves tend to pay fastest. Rewrite the opening of your best-matching page so it answers the question in the first two sentences. Engines cite extractable openings, not buried conclusions. Add a comparison or pricing table where the question implies one; structured contrast is the most-cited format in commercial queries. And check access: a page answer engines cannot fetch (robots rules, JavaScript rendering) earns nothing, however good it is.
From counting citations to acting on them
The count itself is a vanity number until you split it two ways: which of your pages earn citations (double down on that format), and which questions cite competitors but not you (your content gap list, pre-sorted). AI citation analysis walks the full workflow, and the panel's source view builds both lists from your actual scans.