Citation Rank

Citation rank is the position at which a page appears in an AI answer's list of cited sources, counted from 1. If an answer lists five sources and yours is the second, that citation has rank 2. For a brand, citation rank is usually reported as the average rank of all citations to its domain over a period, so a lower number is better.

Source order is not random. The sources an engine lists first are generally the ones its answer relies on most, and they are the ones readers see: a source card at the top of an answer gets attention that the eighth footnote does not. The number of slots is also small. Promptwatch's average sources per response data shows ChatGPT citing about five sources per web-search answer, against roughly ten for Google AI Overviews and Perplexity. Being cited at rank 1 or 2 means your page is a primary source for the answer; rank 9 means you made the bibliography. The original GEO research applied the same logic to visibility by weighting earlier citations more heavily, in its position-adjusted word count metric.

The calculation is simple, but the details change the number:

• Each citation stores its position among that answer's sources, starting at 1.
• Only citations to your own domain feed your rank. Some tools also count subdomains such as docs.yourbrand.com toward the brand.
• Averages are usually computed per day across every citation with a position, so a day with many citations weighs more than a quiet one in a period average.
• Citations returned without a usable order still count as citations but are left out of the rank average.

Citation rank is easy to confuse with answer position. Answer position is where your brand is named among the brands in the answer text; citation rank is where your page sits in the source list. The two often disagree. An answer can recommend you first and cite you last, or use your page as its top source without naming your brand at all. Reading both shows whether your brand and your content are pulling in the same direction.

Rank also has a denominator problem: it only exists where you are cited. Read it next to citation rate and self-citation rate. A worsening average is not always bad news. If you start being cited on prompts where you previously had no citation at all, those new citations often enter at positions 5 to 8 and pull the average up even though your visibility grew. Compare ranks within one engine rather than across engines, because an engine that lists ten sources produces higher averages than one that lists five. Models that answer from training data without searching the web produce no citations and therefore no rank.

Rank improves when models treat your page as the authority for the question rather than as supporting material. Pages that answer the prompt head-on (a pricing question answered with actual prices) tend to be listed before pages that touch the topic in passing. Freshness matters because engines re-rank sources as they re-crawl, so an updated competitor page can push yours down. Competition is per prompt: one strong new page entering the source lists of a high-volume prompt can move your whole average. Tracking cited URL rate and position for individual competitor URLs shows which pages hold the top slots you want, and citations per response tells you how many slots each engine offers.