AI Citation Tracking: Know Which Sources AI Uses for Your Brand
What is AI citation tracking?
AI citation tracking is the practice of identifying which URLs and sources large language models cite — or silently rely on — when they answer questions about your brand. It covers visible citations (the linked sources Perplexity shows) and inferred sources (the URLs ChatGPT, Claude, and Gemini fetched but didn't surface). Citation tracking turns AI from a black box into a debuggable system: every wrong answer can be traced back to a source you can fix, replace, or out-rank.
Why citations matter more than rankings
In SEO, the question is "do I rank?" In AI, the question is "what is AI reading about me?" Two brands can look identical from the outside and get described completely differently by AI — because the citation set behind each answer is different.
Citation tracking is the only honest way to debug an AI answer. Without it, you're guessing why AI gets your category wrong. With it, you can see the exact sources driving the misrepresentation.
59.8% of brand misrepresentation traces back to source disagreement — different platforms describing the same brand differently. Citation tracking is how you find which sources to reconcile first.
What gets cited: the 5 source types
Brand-owned pages
Your homepage, product pages, docs, pricing, glossary. The most controllable source — and the one AI weighs heaviest when it's structured and current.
Structured-data feeds
Your Organization, Product, and SoftwareApplication JSON-LD, plus your llms.txt and ai-agent-manifest.json. AI uses these to disambiguate your entity.
Third-party databases
Crunchbase, Wikipedia, LinkedIn, G2, Capterra. AI treats these as triangulation sources — when they disagree with your site, AI hedges.
Editorial and news
Press, analyst coverage, podcasts, newsletters. Lower frequency but high authority — and the primary driver of recency.
Community and forums
Reddit, Hacker News, Stack Exchange, niche communities. AI increasingly cites these for buyer-experience questions ('is X actually good for Y').
How to track citations across ChatGPT, Claude, Gemini, and Perplexity
Perplexity exposes citations directly — every answer lists its sources. This is the easiest model to audit.
ChatGPT and Gemini surface citations when they browse, but hide parametric sources. To audit, ask the model to "list the URLs you would check to answer this question about [brand]" and cross-check against your server logs.
Claude rarely surfaces URLs unless explicitly asked. Server logs are the only reliable signal — look for ClaudeBot, OAI-SearchBot, GPTBot, and PerplexityBot in your access logs after a query is run.
Cross-reference with AI brand visibility tracking so you can see citation patterns over time, not just point-in-time snapshots.
Owned, earned, and third-party citations
Owned citations are brand-controlled URLs AI cites: your site, docs, glossary, primary source. Maximum control, and AI weighs them heavily when they're structured and current.
Earned citations are URLs you influenced but don't own: guest posts, podcast transcripts, analyst notes. Lower volume, higher authority signal.
Third-party citations are everything else: reviews, news, databases, forums. You can't control them, but you can monitor them and intervene when they're wrong about you.