Understand citation rate

Citation rate helps you assess references to your own website in collected AI answers. Use it alongside brand mentions to understand where the answer points readers for supporting information.

How Erlin calculates it

For every answer, Erlin divides the sources on your website's domain by all the sources the answer cites. An answer that cites no sources counts as 0%. Citation rate is the average of those shares across every answer in the range.

Citation rate = average of (your site's sources ÷ all sources cited) × 100, over every answer

Worked example: Acme has 4 answers. They cite 5 sources (1 is Acme's), 4 sources (2 are Acme's), 6 sources (none) and no sources. The shares are 20%, 50%, 0% and 0%: (20 + 50 + 0 + 0) ÷ 4 = 17.5%, shown as 18%.

Read the metric in context

Open AI Visibility > Overview and find Citation rate in Visibility over time. Select the reporting period and review the prompt scope before comparing results.

A brand can be named while the answer cites only third-party pages. Your website can also be cited without the business being recommended. Neither event proves that someone clicked the link.

Check the answers behind the number

Open a relevant prompt in Prompt details. Review its platform, answer and sources. Compare like-for-like questions before attributing a movement to a content change.

Open the cited page and check which statement it supports. A source about your category is different from a page that accurately describes your own service.

Avoid misleading comparisons

Changing the prompt set, platforms or the locations prompts run from changes the evidence being summarized. Keep a record of those changes. A period with missing answers should not be treated as equivalent to a complete set of checks.

Prompt Discovery volume is an estimate of demand. It is separate from the answers used to assess visibility, and neither number is a count of website visitors.

Decide what to investigate next

Review competitor gaps and cited sources to find the specific page or question behind a change. Use Brand mention rules for recognition errors, and recommendations for possible next actions.

See How Erlin works for the measurement boundaries. A stronger metric is evidence about tracked answers, not a guaranteed business outcome.