# Citation Rate

Citation Rate is the ratio of eligible AI answer runs that cite a specified domain or entity to all eligible runs under a declared query set, engine set, matching rule, and observation window.

Canonical URL: https://machinerelations.ai/glossary/citation-rate
Category: metrics

## Source Body

> Premise: **Citation Rate is the percentage of eligible AI answer runs that cite a specified domain or entity under a declared query set, engine set, matching rule, and observation window.** It measures source-selection frequency. It does not, by itself, measure mentions, recommendations, traffic, revenue, or overall AI visibility.

## Citation Rate formula

Premise: **The basic Citation Rate formula uses answer runs as the unit of observation:**

> Arithmetic illustration: **Citation Rate = (eligible answer runs with at least one qualifying citation ÷ all eligible answer runs) × 100**

Premise: Count each answer run once in the numerator, even if the answer links to the same measured domain several times. The denominator must include every eligible run in the declared panel, including runs with no citation to the measured domain. A run should be excluded only under a rule defined before collection, such as an engine error or an answer that was not generated.

Google describes its AI search features as producing responses with supporting web links and using query fan-out across related searches and data sources ([Google Search Central](https://developers.google.com/search/docs/appearance/ai-features)). Google's official optimization guide also describes AI Overviews and AI Mode as generative search experiences that help people discover relevant websites ([Google Search Central](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)).

Premise: Citation Rate turns those observable source selections into a repeatable frequency measure.

## Citation Rate denominator choices

Premise: **A Citation Rate is interpretable only when its denominator and counting unit are named.** Different denominator choices answer different questions and should not share the same unlabeled percentage.

| Denominator | Formula unit | Question answered | Main caveat |
|---|---|---|---|
| Premise: Answer runs | Premise: Runs citing the entity ÷ all eligible runs | Premise: How often is this entity cited when an answer is generated? | Premise: Repeated runs and engine mix must be declared. |
| Premise: Prompts | Premise: Prompts with at least one citing run ÷ all tracked prompts | Premise: Across how much of the prompt set does the entity appear? | Premise: Requires a fixed rule for combining multiple engines or repeated runs. |
| Premise: Citation opportunities | Premise: Qualifying citations to the entity ÷ all qualifying citations | Premise: What share of citation slots does the entity receive? | Premise: This is closer to share of citation than run-level Citation Rate. |
| Premise: Answers with any citation | Premise: Runs citing the entity ÷ runs containing at least one citation | Premise: How often does the entity appear when the engine cites anything? | Premise: Excluding uncited answers can make the rate look higher. |

Premise: Machine Relations uses eligible answer runs as the default denominator for Citation Rate because the unit stays tied to a generated answer. Prompt coverage and citation-opportunity share remain useful, but they should be labeled separately rather than blended into one metric.

## Worked Citation Rate example

Premise: **A hypothetical panel with 120 eligible answer runs and 18 runs citing the measured domain has a Citation Rate of 15%.**

> Premise: **18 ÷ 120 × 100 = 15%**

Premise: Suppose the panel contains 20 buyer questions tested on three AI engines, with two runs per question-engine pair. That produces 120 planned runs. If four runs fail and the predeclared protocol excludes engine failures, the eligible denominator becomes 116 rather than 120. If 18 of those eligible runs cite the domain, the reported rate is 15.5%, calculated as 18 ÷ 116 × 100.

Premise: The report should show the numerator and denominator beside the percentage: **18 cited runs / 116 eligible runs = 15.5%**. This is an arithmetic example, not an industry benchmark or a claim about what rate is good.

## Citation Rate vs share of citation, mention rate, and visibility score

Premise: **Citation Rate measures source selection; adjacent metrics measure competitive share, textual presence, or a composite view.** Treating them as interchangeable hides the behavior each metric is meant to observe.

| Metric | What it counts | Typical denominator | What it does not prove |
|---|---|---|---|
| Premise: **Citation Rate** | Premise: Answer runs containing a qualifying citation to a domain or entity | Premise: All eligible answer runs in the declared panel | Premise: Recommendation, sentiment, clicks, or business impact |
| Premise: **[Share of Citation](https://machinerelations.ai/glossary/share-of-citation)** | Premise: The focal entity's qualifying citation presence relative to a declared citation or competitor set | Premise: All qualifying citations or citation presence in that set | Premise: Absolute coverage across all tracked runs |
| Premise: **Mention rate** | Premise: Answer runs that name the entity, whether linked or not | Premise: All eligible answer runs | Premise: That the engine used the entity as a source |
| Premise: **[AI Visibility Score](https://machinerelations.ai/glossary/ai-visibility-score)** | Premise: A declared combination of signals such as citations, mentions, rankings, sentiment, or recommendations | Premise: The scoring model's normalized scale | Premise: A directly observed event unless its components are disclosed |

Premise: A brand can have a high mention rate and a low Citation Rate when engines discuss the brand but cite other sources. It can also have a strong Citation Rate and weak recommendation presence when its pages are used for facts but the brand is not presented as the preferred choice.

Microsoft's first-party AI reporting separates related concepts including citation share, cited pages, and grounding queries rather than collapsing them into one number ([Microsoft Bing, 2026](https://blogs.bing.com/search/June-2026/New-AI-Visibility-Insights-in-Bing-Webmaster-Tools-Intents-Topics-Citation-Share-Compare)).

Premise: The labels vary by platform, so the underlying numerator, denominator, and matching rule matter more than the dashboard name.

## Citation Rate measurement contract and caveats

Premise: **A defensible Citation Rate requires a fixed measurement contract before collection begins.** At minimum, declare:

1. Premise: the domain, URL, brand, or entity being matched;
2. Premise: the exact query or prompt set and its segment labels;
3. Premise: the AI engines and product surfaces included;
4. Premise: the number of repeated runs per prompt and engine;
5. Premise: the start and end dates and collection cadence;
6. Premise: what qualifies as a citation;
7. Premise: how redirects, subdomains, syndicated URLs, and third-party articles are matched;
8. Premise: the exclusion rule for failed or incomplete runs; and
9. Premise: the numerator, denominator, and sample size reported with the percentage.

Premise: Citation Rate can vary when prompts, engines, model versions, locations, personalization, or observation windows change. Rates from two vendors or two internal panels are not directly comparable unless their contracts match. Repeated observations can describe change within a stable panel, but they do not establish why the change occurred.

Premise: There is no universal good Citation Rate. A useful baseline is the rate measured repeatedly for the same entity, segment, engines, and protocol, alongside relevant competitors or source types. Small samples should be reported as provisional rather than converted into false precision.

## Citation Rate in Machine Relations

Premise: **Within [Machine Relations](https://machinerelations.ai/glossary/machine-relations), Citation Rate is a measurement-layer diagnostic for whether machines select a brand's evidence as source material.** It isolates a source-selection observation that operators can evaluate alongside earned authority, entity clarity, citation architecture, distribution, and business outcomes.

Premise: Citation Rate does not explain the whole system. It supplies a concrete observation that operators can investigate. If a segment's rate is weak, the next questions concern source authority, entity matching, extractability, coverage across trusted publications, and whether the measured content actually answers the tracked prompts.

## Frequently asked questions about Citation Rate

**How do you calculate Citation Rate?**

Premise: Divide eligible answer runs containing at least one qualifying citation to the measured domain or entity by all eligible answer runs, then multiply by 100. Report the numerator and denominator with the percentage.

**What should count as a citation?**

Premise: Use a predeclared observable rule, such as a visible source link or source-card URL that resolves to the measured domain. Do not silently count an unlinked brand mention as a citation.

**What is the difference between Citation Rate and mention rate?**

Premise: Citation Rate counts qualifying source attributions; mention rate counts answers that name the entity. A mention can occur without a citation, and a citation can point to evidence about an entity without recommending it.

**What is the difference between Citation Rate and Share of Citation?**

Premise: Citation Rate asks how often a source appears across eligible answer runs. Share of Citation asks how much of a declared competitive citation set belongs to that source. The former is an absolute run-frequency measure; the latter is a relative share measure.

**What is a good Citation Rate?**

Premise: No universal benchmark is defensible across different query sets, engines, dates, and matching rules. Compare like with like: the same panel over time or entities measured under the same contract.

## Sources

- https://developers.google.com/search/docs/appearance/ai-features
- https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- https://blogs.bing.com/search/June-2026/New-AI-Visibility-Insights-in-Bing-Webmaster-Tools-Intents-Topics-Citation-Share-Compare

## Machine Relations references

Machine Relations' own methodology, dataset, and research pages related to this term. These are self-references, listed separately from Sources — they are not independent evidence.

- https://machinerelations.ai/glossary/share-of-citation
- https://machinerelations.ai/glossary/ai-visibility-score
- https://machinerelations.ai/glossary/machine-relations

## Machine-readable related links

### Related concepts

- [Share of Citation](https://machinerelations.ai/glossary/share-of-citation)
- [Machine Relations (MR)](https://machinerelations.ai/glossary/machine-relations)
- [AI Visibility](https://machinerelations.ai/glossary/ai-visibility)
- [Machine Relations Index (MRI)](https://machinerelations.ai/glossary/machine-relations-index)

### Supporting research

- [Crunchbase Answer-Engine Citation Authority: 2.76% Citation Rate, B-Confidence — Claude Emerges as Dominant Engine](https://machinerelations.ai/research/crunchbase-answer-engine-citation-authority-2026)
- [Entity Chain Architecture: How Brands Build Linked Proof Networks That AI Engines Actually Cite](https://machinerelations.ai/research/entity-chain-architecture-linked-proof-networks-ai-citations-2026)
- [Entity Chain Requirements by AI Platform: What ChatGPT, Perplexity, and Gemini Need to Cite Your Brand](https://machinerelations.ai/research/entity-chain-requirements-by-ai-platform-citation-2026)
- [Fortune Business Insights Answer-Engine Citation Authority: 1.58% Citation Rate Across 16,898 Domains — Market Sizing Infrastructure Holds](https://machinerelations.ai/research/fortunebusinessinsights-answer-engine-citation-authority-mri)

### Framework context

- [Machine Relations Index](https://machinerelations.ai/index)
- [Machine Relations Stack](https://machinerelations.ai/stack)
- [Evidence Base](https://machinerelations.ai/evidence)