What is the difference between mention rate and citation rate in GEO?
Mention rate measures how often a business is named in a defined set of AI answers. An answer-based citation rate measures how often that answer set references a defined source, such as your domain. A mention need not contain a citation, and a citation need not name the brand. Always state the denominator: 3LA Search Insights uses a different, cross-report citation-to-search-impressions ratio.
Define the answer set before calculating a rate
For a spontaneous brand-visibility test, use questions that do not name the business. A question such as “What does [brand] offer?” already supplies the name and should not be mixed into that mention-rate denominator. Named questions can still be useful for checking reputation or factual accuracy; report them separately.
Record the questions, AI service, model or surface, language, location, browsing mode, dates and valid run count. If you compare mention rate with answer-based citation rate, use the same eligible answer set. When source reporting is unavailable, report citation rate as unavailable rather than silently counting a missing field as zero.
- Mention rate = eligible answers naming the business ÷ eligible answers × 100.
- Answer-based citation rate = eligible answers with at least one source reference to the defined target ÷ eligible answers × 100.
- Count each answer at most once per metric, even when it names the brand or links to the domain several times.
A worked example: two different signals
Illustrative data, not a customer result: take 50 valid answers to non-branded questions from a single defined test. The business is named in 20 answers; 8 answers link to its domain. Mention rate is 20 ÷ 50 = 40%. Answer-based citation rate is 8 ÷ 50 = 16%.
The sets may overlap, so do not add 40% and 16%. An answer naming the business while linking to an independent publisher contributes to the mention rate but not to the domain-citation rate. If you also track third-party references about the business, define that as a separate measure.
Illustration: answers citing your domain
- Answers with a domain citation
- 8
- Eligible answers
- 50
- × 100 · answer-based citation rate
- 16%
3LA Search Insights uses a different denominator
In 3LA Search Insights, Citation Rate (Bing+GSC) is reported AI citations divided by combined Bing and Google search impressions, multiplied by 100. It combines reports rather than counting a single set of observed AI answers. It is not a probability that a search impression becomes an AI citation.
The research variant CFR (GSC-only) uses Google search impressions only. Neither of these ratios is interchangeable with the answer-based 16% in the example. Keep the metric name, sources, page set and compatible reporting periods with the number.
The product displays no rate below 100 search impressions and caps displayed rates at 100%. The underlying cross-source ratio can mathematically exceed 100 because the sources count different events. Google AI impressions are not Bing AI citations and must not be added on top of total Google search impressions.
What should you do with the result?
Compare like with like over time and report each service separately. A change from 8 to 10 cited answers may be worth following, but is not proof of improvement without considering sample size and variability. Track clicks, contact goals and qualified enquiries separately: neither rate is a lead or sales conversion rate.
- High mention rate, low domain-citation rate: inspect which publishers supply the evidence and whether your pages provide a usable original source.
- Domain citations without brand mentions: check how the answer attributes the source and whether your entity is clear on the cited page.
- Low rates for both: check question relevance, content access and evidence before producing more content.
- Small or changing samples: repeat the observations and show counts; do not interpret one answer as a trend.
Frequently asked questions
Is there a universally good citation rate?
No. The denominator, AI service, question set, citation target and period determine what the rate means. Compare a stable measurement with its own baseline rather than using an unrelated benchmark.
Does a missing citation field mean the answer had no sources?
Not necessarily. Your tool or service may not expose source references. Mark that coverage as unavailable and distinguish it from a valid observation where source references were inspected and none matched.
Is Citation Share the same as citation rate?
Not automatically. For example, Bing Citation Share concerns a site's share of citations for a grounding query. The numerator and denominator differ from both an answer-sample rate and 3LA's citation-to-search-impressions ratio. Check the provider's definition before comparing.
Sources and scope
The answer-based definitions are explicitly scoped to the observation method described here; tools can use different definitions. The worked example is illustrative. Product-specific definitions follow the existing 3LA GEO guide.
Agree on the denominator first
Before setting a target, choose the question set, citation target and reporting period. If your reports use conflicting definitions, we can help you separate the measurements.
Talk to Nettpilot about where to start →
