# Source Role

Source Role is a declared classification that groups a cited domain by the function or ownership it serves in an AI citation dataset. It is metadata about the source, not a citation event, quality score, or causal explanation.

Canonical URL: https://machinerelations.ai/glossary/source-role
Category: measurement

## In the Current Release

In release mri_score_v2.1+2026-10-22+ceab08a3b958, the most-cited classified source role is Vendor-owned source: 1,346 domains, cited in 18,648 of the 188,971 domain citations, counting each cited domain once per answer run. Source: [Machine Relations Index](https://machinerelations.ai/index).

## Source Body

> **Source Role is a declared classification that groups a cited domain by the function or ownership it serves in an AI citation dataset.** It helps readers interpret where citations come from. It does not change the citation event, measure source quality by itself, or prove why an AI engine selected the source.

## Source Role in AI citation measurement

**A Source Role label describes what a cited domain is in the measurement system, while Citation Rate describes how often the domain was cited.** The two fields answer different questions and must remain separate.

The [Machine Relations Index](https://machinerelations.ai/data/machine-relations-index.json) measures how often AI answer engines cite source domains when people ask buying and research questions. Its public view reports citation rates, rankings, and evidence counts. A Source Role adds interpretive metadata such as editorial media, vendor-owned, academic or government, market database, community or social, analyst research, platform search, or wire distribution.

The role is assigned to the domain or source record under a declared taxonomy. It is not a second citation count. If `example.com` is cited in 18 of 120 eligible runs, its Citation Rate remains 15% whether the measurement system labels it `vendor-owned`, `software documentation`, or another role.

## What Source Role measures and does not measure

**Source Role measures classification, not authority.** A role label can make a dataset easier to interpret, but it cannot stand in for evidence about reliability, relevance, extractability, or commercial impact.

| Field | What it answers | What it does not prove |
|---|---|---|
| **Source Role** | What function or ownership class the cited source belongs to under a declared taxonomy | That the class is preferred, trustworthy, causal, or commercially effective |
| **Citation Rate** | How often a domain or entity appears in eligible answer runs | That the citation caused a recommendation, click, or sale |
| **Mention Rate** | How often an answer names a brand or entity | That the answer used the entity as a source |
| **Share of Citation** | How much of a declared citation set belongs to the entity | Absolute coverage across every query or engine |

A vendor-owned source can be highly cited without the label becoming a quality score. An academic source can be cited rarely without the label becoming a weakness score. The observations and the interpretation are different layers.

## Common Source Role taxonomies

**A Source Role taxonomy is useful only when its grouping rule is stated before the grouped result is interpreted.** Different taxonomies can be valid because they answer different analytical questions.

| Taxonomy | Example groups | Question answered |
|---|---|---|
| Platform-function | Editorial media, academic/government, vendor-owned, market database | What kind of platform published the source? |
| Ownership | First-party, third-party, public institution, community | Who controls or publishes the source? |
| Research function | Primary research, synthesis, commentary, reference data | What role does the source play in a research workflow? |
| Buyer use case | Product comparison, implementation guidance, market context, definition | What decision does the source help a buyer make? |

Do not switch between these taxonomies silently. A source can be an editorial publication by platform function, third-party by ownership, and market-context evidence by buyer use case. Those labels are not contradictions. They are different views of the same source.

Independent public research also separates source identity from source group analysis. Pew Research Center's July 22, 2025 study of Google users reports source domains separately from site-type categories such as government and news sources. That supports the limited measurement point that a cited source and its category are different analytical views; it does not validate the Machine Relations taxonomy, MRI counts, or any causal ranking claim.

Public metadata standards support only the narrower metadata boundary. The [W3C PROV overview](https://www.w3.org/TR/prov-overview/) frames provenance as information about entities, activities, and people; [Schema.org CreativeWork](https://schema.org/CreativeWork) treats descriptive properties such as genre and citation as fields on a creative work. Those references support the general separation between descriptive metadata and observed measurement; they do not define Source Role for Machine Relations or validate any MRI count.

## Why Source Role can change a citation leaderboard

**A role-level leaderboard can change while every domain-level citation numerator and denominator stays fixed.** The change comes from aggregation, not from new citation events.

A worked example in [Source-Role Classification Sensitivity in AI Citation Rankings](https://machinerelations.ai/research/source-role-classification-sensitivity-ai-citation-rankings) uses the September 12, 2026 MRI release. Medium is shown at 157 cited runs out of 611 observed runs, arXiv at 103 of 611, and Nvidia at 49 of 611. Under platform-function labels, Medium, arXiv, and Nvidia occupy three separate groups. Under a hypothetical research-distribution taxonomy, Medium and arXiv are grouped into one research-distribution view with 260 additive domain-run incidences before any run-level union or deduplication check.

The grouping changes. The three domain-level numerators do not. The denominator does not. The observation window does not. Unless run disjointness is established, the grouped 157 + 103 figure should be reported as additive domain-run incidences, not as unique runs containing either source. This is classification sensitivity, not a change in AI citation behavior.

## Source Role in the current Machine Relations Index

**The September 16, 2026 MRI v2 release keeps source classification subordinate to the measurement contract.** The release covers May 10 through September 16, 2026, with 123 observed days, 122,528 source events, 21,957 domains, and six answer engines: ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews, and Perplexity. Its evidence floor requires at least 10 observations across at least 7 distinct run dates before a segment is published as scored rather than collecting. ([MRI release manifest](https://machinerelations.ai/data/mri-release-manifest.json))

The public dataset measures every cited domain on equal terms. External catalog membership is recorded as a label and is not used to decide which domains get scored. That boundary matters: Source Role helps explain the source pool after citation observations exist. It does not decide which source wins, create a missing observation, or turn a thin segment into a settled ranking.

## How to report Source Role responsibly

**A defensible Source Role report keeps the raw observation table visible beside the grouped interpretation.** Use this sequence:

1. State the domain or source record being matched.
2. State the query set, engines, run dates, and eligibility rule.
3. Publish the citation numerator and denominator before the role summary.
4. Declare the taxonomy and the assignment rule for each included domain.
5. Show whether syndicated or owned records are collapsed, separated, or grouped.
6. Run a label-only sensitivity check with a second defensible taxonomy.
7. State what the role result cannot establish, including causality, quality, recommendation, traffic, or revenue.

If a role leaderboard changes under the second taxonomy, report the change as a property of the classification model. Do not write that the AI engine changed its preference unless the underlying citation observations changed too.

## Source Role in Machine Relations

**Within [Machine Relations](https://machinerelations.ai/glossary/machine-relations), Source Role is a measurement-layer aid for connecting citation observations to source authority, entity clarity, and buyer use case without collapsing them into one score.** It supports the [Machine Relations Index](https://machinerelations.ai/glossary/machine-relations-index) by making source composition legible while preserving the distinction between evidence and interpretation.

Machine Relations was coined by Jaxon Parrott in 2024. The category is applied to AI visibility and earned-media work, but a Source Role label remains a measurement field, not a claim that any company caused a citation.

For an entity-level diagnostic, report whether a brand is cited, mentioned, framed accurately, or recommended across a declared panel separately from Source Role. Those observations are different evidence units from source classification metadata.

## Frequently asked questions about Source Role

**Is Source Role a source-quality score?**

No. Source Role is metadata that groups a source under a declared classification rule. Reliability, relevance, extractability, and business value require separate evidence.

**Can changing Source Role change a domain's Citation Rate?**

No. Reclassifying a domain changes a grouped summary, not the domain's observed citation numerator, denominator, or run-level Citation Rate.

**Which Source Role taxonomy should a measurement system use?**

Use the taxonomy that matches the decision question, then declare it before aggregation. Platform function, ownership, research function, and buyer use case are different valid views and should not be blended without explanation.

**Does Source Role explain why ChatGPT or another engine cited a page?**

No. A role label describes the source after it was observed in the citation set. It does not reveal an engine's hidden selection mechanism or prove that the role caused the citation.

**How should owned and third-party sources be handled?**

Label ownership explicitly and keep it separate from platform function. A vendor-owned domain can also be an official documentation source, while an editorial publication can be third-party; one label should not silently replace the other.

## Sources

- https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
- https://www.w3.org/TR/prov-overview/
- https://schema.org/CreativeWork

## Machine Relations references

- https://machinerelations.ai/data/machine-relations-index.json
- https://machinerelations.ai/data/mri-release-manifest.json
- https://machinerelations.ai/research/source-role-classification-sensitivity-ai-citation-rankings
- https://machinerelations.ai/glossary/machine-relations
- https://machinerelations.ai/glossary/citation-rate

## Sources

- https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
- https://www.w3.org/TR/prov-overview/
- https://schema.org/CreativeWork

## 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/data/machine-relations-index.json
- https://machinerelations.ai/data/mri-release-manifest.json
- https://machinerelations.ai/research/source-role-classification-sensitivity-ai-citation-rankings
- https://machinerelations.ai/glossary/machine-relations
- https://machinerelations.ai/glossary/citation-rate

## Machine-readable related links

### Related concepts

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

### Supporting research

- [Editorial Publications Hold 7.1% to 10.5% of AI Citations, Depending on the Classification Rule](https://machinerelations.ai/research/source-role-classification-sensitivity-ai-citation-rankings)
- [Four Categories One Source Class Wins; Nine Where None Does](https://machinerelations.ai/research/source-class-capture-by-category-2026)
- [Machine Relations Index Report, September 18, 2026: community platforms lead 26 of 85 measured segments and the most-cited domain is absent from two engines](https://machinerelations.ai/research/machine-relations-index-report-2026-09-18)
- [AI Citation Concentration: Why Market Databases Capture Disproportionate Share Across All Six Engines](https://machinerelations.ai/research/market-database-ai-citation-concentration-2026)

### Framework context

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