Definition
Citation rate is the share of tracked prompts where an AI engine links to your domain as a source. The formula is simple—responses citing your domain divided by total responses in the prompt set, expressed as a percentage—but it answers a specific question: how often does this engine reach for my content at all?
It is easy to confuse with citation share, and the difference matters. Citation rate is absolute: out of 100 prompts, how many cited me? Citation share is relative: of all citations across a competitor set, what proportion were mine? A niche brand can hold high citation share on the handful of prompts where it appears while running a low citation rate overall. Reporting one without the other consistently misleads.
Citation rate is also distinct from mention rate, which counts brand-name appearances whether or not a link is attached. The gap between them is the ghost citation rate, and tracking both is what keeps AI visibility reporting reconcilable with referral analytics.
Measurement discipline matters more here than for traditional rank tracking. AI answers are non-deterministic, so the same prompt can cite different sources across runs, and citation volume differs sharply by engine—analyses have found Perplexity averaging well over twenty citations per answer against single digits for ChatGPT. That means raw rates are not comparable across engines. Freeze the prompt set, sample each prompt several times per engine per cycle, and report per engine rather than blending everything into a single number.
Examples of Citation Rate
- A brand cited in 18 of 200 monitored prompts reports an 18/200 = 9% citation rate for that engine and period.
- A team reports citation rate per engine after finding their Perplexity rate is several times their ChatGPT rate purely because Perplexity cites more sources per answer.
- A niche vendor with 40% citation share on its narrow prompt set discovers its overall citation rate is only 4%, revealing how few prompts it competes on.
- An analyst runs each prompt three times per week and averages the results, because single measurements swung by double digits between runs.
