Citation rate

Long definition

Citation rate answers a specific question: out of N AI answers generated for queries you care about, in how many did the model link to or attribute your domain? You pick a query universe (your tracked keywords, your topic clusters, brand+category prompts), run them across one or more AI surfaces, and count citations. Divide by N and you have your rate for that engine, that day.

The metric is young. There is no Search Console for AI citations, no single tool everyone agrees on. Vendors like Profound, Otterly, Peec, Daydream, and Athena all run their own probe sets against ChatGPT, Perplexity, Gemini, and Claude — results diverge because the underlying retrieval is non-deterministic and personalized. A citation rate of 12% on Profound and 7% on Otterly for the same domain is normal noise.

What makes the number useful is the trend, not the absolute. A site that goes from 4% to 11% over a quarter on a stable query set is gaining ground in the model's retrieval index. A site whose rate collapses after a Common Crawl refresh has lost its place in training or live retrieval data.

Treat citation rate as a directional KPI, not a precise measurement. Pair it with co-citation analysis (which other domains are cited alongside you), citation position (first, middle, last in the answer), and query coverage (which slices of your topic map are blank). One number per quarter beats one number per day if your query set is too small.

Common misconceptions

• "Citation rate is the AI version of organic rankings." Closer to share-of-voice across an answer set. There's no position 1 — there's "cited or not cited" plus how prominently. A page can rank #1 organically and still have 0% citation rate if the model never retrieves it.
• "More content = higher citation rate." Volume without retrievability does nothing. Models cite content they can find, parse, and verify. Pages behind login walls, JavaScript-only renders, or buried 6 clicks deep rarely get cited regardless of word count.
• "You can A/B test citation rate." Not reliably. The same prompt run twice can return different sources. Sample sizes need to be large (hundreds of queries, repeated runs) before differences become signal.
• "Citations always link out." Many AI surfaces show citations as inline numerals or sidebar tiles. Click-through is dramatically lower than from a blue link. Citation rate is a visibility metric, not a traffic metric.