AI citation rate is the percentage of relevant buying queries where an AI assistant names or cites your company. Ask the twenty questions a buyer in your category actually asks — *best X for Y*, *how much does X cost*, *X vs Y* — across ChatGPT, Perplexity, and Google AI Overviews. If you appear in the answer for six of twenty, your citation rate is 30%. It's the closest thing GEO has to rank tracking, and it's the number a serious program is accountable to.
Measured honestly, it needs three controls. A fixed query set — defined from real buyer language, not vanity prompts you already win. Multiple engines — each assistant retrieves differently; ChatGPT leans on Bing's index, AI Overviews on Google's, Perplexity on its own crawler, and a brand routinely appears in one while invisible in another. Repetition over time — answers are stochastic, so a single run is an anecdote; run the panel monthly and track the trend, not the day.
What's *good* depends on where you start. A brand with no GEO work typically measures 0–10% — invisible outside its own name searches. A deliberate program reaching 30–40% on commercial queries is winning its category; past 50%, you're the default answer competitors have to displace. The more useful number early on is competitive share: on queries where *someone* in your category is named, how often is it you?
The levers that move it are knowable: pages that answer the exact query (not adjacent ones), quotable facts like published pricing, FAQ and schema markup that survive extraction, and entity consistency the engines can verify. We covered the mechanics in how assistants choose who to recommend and the full system in the GEO field guide.
Two failure modes to avoid. Don't measure only your branded queries — 'what is [your company]' is recognition, not acquisition; the money is in category queries where buyers don't know you yet. And don't chase citations on queries with no commercial intent; a citation on *'history of automation'* is trivia, while a citation on *'how much does an AI agent cost'* is pipeline.
If you want the baseline without building the measurement harness yourself, the free citation audit runs your category's queries and sends you the scored result — where you show up, where you don't, and who's being named instead. That report is also exactly how GEO-as-a-Service is scoped and, later, how it's held accountable: the retainer is measured on this number.
