How to track AI citations (and why it matters more than rank)

Rank tracking measures: did Google list me in the top 10 for this keyword?

Citation tracking measures: did ChatGPT (or Perplexity, or Gemini, or Google's AI Overview) cite my site as the answer?

These are different questions, and citation tracking matters more — because in an AI-search world, being on page 1 of Google means almost nothing if the AI Overview at the top doesn't cite you.

How to measure citations

Pick five representative queries, run each through ChatGPT, Perplexity, Gemini, and Google's AI Overview on the same day each week, and score binary — cited or not. Track the total as a fraction: three out of five cited is your North Star. "Mentioned without a source" does not count as a citation.

1. Pick 5 priority queries. Brand, "best in city," "near me," a long-tail service question, a pain-point question. The same 5 queries every week.
2. Run each query through each engine. Programmatically (we use Perplexity's sonar-pro model with return_citations=true); manually if you're starting out (paste into ChatGPT, screenshot the sources).
3. Score binary: cited or not. "Mentioned without source" is not cited. Be strict.
4. Track the count weekly. X / 5 cited is your North Star number.

What moves the number

Schema markup is the fastest lever — LocalBusiness and Service structured data give AI engines a structured signal to cite you instead of aggregators like Yelp. Answer-first articles on the exact query phrasing, an active GBP, and fresh reviews compound over weeks to lock in your citation position.

In rough order of impact:

1. LocalBusiness + Service schema sitewide. Without it, AI engines fall back to aggregators like Yelp.
2. Answer-first articles on the exact query phrasing. "Best [service] in [city]" → write an article called "How to choose the best [service] in [city]."
3. GBP active. AI engines pull from local-pack signals.
4. Reviews on Google + Yelp + your own page. AI engines factor reputation.
5. Fresh content recency. Two articles a week beats four a month.

The compounding effect

Citation rate is slow to move (weeks, not days) and slow to lose (weeks of inactivity before it slips). Once you cross 3/5 cited, you usually stay there — even if you take a month off — because the AI engines have crystallized your brand as the answer.

This is why AEO is a moat, not a project.