---
title: "Most Cited Sources in Google AI Mode: Citing Answers Name a Median of Three and Two Sources Lead"
description: "Since August 31, Google AI Mode answers that cite a source name a median of 3; across 908 runs the youtube.com bucket of unmapped YouTube pages and Reddit lead."
canonical: https://authoritytech.io/blog/top-google-ai-mode-citation-domains-machine-relations-index-2026
last-updated: 2026-11-05
---

# Most Cited Sources in Google AI Mode: Citing Answers Name a Median of Three and Two Sources Lead

Since August 31, Google AI Mode answers that cite a source name a median of 3; across 908 runs the youtube.com bucket of unmapped YouTube pages and Reddit lead.

Canonical URL: https://authoritytech.io/blog/top-google-ai-mode-citation-domains-machine-relations-index-2026
Published: 2026-11-05
Author: Jaxon Parrott
Topic: Machine Relations

<p><strong>In the Machine Relations Index release of November 5, 2026, Google AI Mode answers collected since August 31 cite few sources. Of 908 observed runs, 606 (66.7%) cited at least one source, the median run that cited any cited 3 different sources, and the answers cited 774 different sources in all. Only two were cited in more than 2% of runs: the youtube.com bucket, which holds only YouTube pages not mapped to a channel or publication (98 runs, 10.79%), and Reddit (76 runs, 8.37%).</strong></p>

<p>That is the answer to "which sources does Google AI Mode cite most," and it changes how to read a top-sources list. An AI Mode answer names a handful of sources, a third of the Index's runs named none, and the most-cited source appears in about one run in nine. The Index is a first-party dataset, published by Machine Relations Research and created by Jaxon Parrott, who founded AuthorityTech; it is not an independent audit. It measures which sources the answers cite in the Index's monitored question libraries. It does not measure buyer demand, and it does not show that any placement caused a citation.</p>

<h2>Key takeaways</h2>
<ul>
<li><strong>Few sources per answer.</strong> Of 908 observed runs since August 31, 606 cited at least one source and the median citing run cited 3. The other 302 runs cited none, and they still count in the 908.</li>
<li><strong>A short head and a long tail.</strong> The answers cited 774 different sources. Only the youtube.com bucket (10.79%) and Reddit (8.37%) were cited in more than 2% of runs; the eighth source, Forbes, was cited in 10 runs (1.10%).</li>
<li><strong>Publications are small slices.</strong> The New York Times (15 runs, 1.65%) and Forbes (10 runs, 1.10%) are the editorial publications among the first eight; Reddit, LinkedIn and Medium are the community and social platforms.</li>
<li><strong>youtube.com is a bucket.</strong> It counts only YouTube pages the Index could not map to a channel or publication, so it is not a count of all YouTube citations.</li>
</ul>

<h2>Which sources does Google AI Mode cite most?</h2>

<p>The table lists the sources cited in the largest share of the 908 Google AI Mode runs the Index has observed since August 31, with the Index's role for each source. Ranks are the Index's, and tied sources share a rank.</p>

<table>
<thead><tr><th>Rank</th><th>Source</th><th>Source role in the Index</th><th>Runs citing it</th><th>Share of 908 runs</th></tr></thead>
<tbody>
<tr><td>1</td><td>youtube.com (YouTube pages not mapped to a channel or publication)</td><td>Search or media platform</td><td>98</td><td>10.79%</td></tr>
<tr><td>2</td><td>Reddit</td><td>Community and social platform</td><td>76</td><td>8.37%</td></tr>
<tr><td>3</td><td>Gartner</td><td>Analyst and consulting research</td><td>17</td><td>1.87%</td></tr>
<tr><td>4</td><td>cdp.com</td><td>Vendor-owned source</td><td>15</td><td>1.65%</td></tr>
<tr><td>4</td><td>LinkedIn</td><td>Community and social platform</td><td>15</td><td>1.65%</td></tr>
<tr><td>4</td><td>The New York Times</td><td>Editorial publication</td><td>15</td><td>1.65%</td></tr>
<tr><td>7</td><td>Medium</td><td>Community and social platform</td><td>14</td><td>1.54%</td></tr>
<tr><td>8</td><td>Forbes</td><td>Editorial publication</td><td>10</td><td>1.10%</td></tr>
</tbody>
</table>

<p>Beyond the first eight sources, no source was cited in more than 8 of the 908 runs (0.88%), so everything below the table is a long tail of single-digit counts, not a ranking worth reading closely.</p>

<p>One unit to read first: the Index counts resolved sources. A YouTube page it can map to a channel or publication is credited to that channel or publication, so youtube.com here is only the YouTube pages it could not map. It is a bucket, not the total of YouTube citations, and it should not be read as "YouTube" the platform.</p>

<h2>Why this page uses only the data since August 31</h2>

<p>Google AI Mode collection changed between August 10 and August 30, 2026, and answers collected after the change cite far fewer distinct sources. The Index therefore reports the period from August 31 as one collection regime, on its own. A ranking that pools the months before and after the change describes a mixture of two collection methods and does not tell you who is cited in AI Mode answers today. The Index says the two periods are not comparable, do not form a trend, and that neither shows its collection captures everything Google AI Mode shows a person. This page makes no comparison with earlier months, and it does not claim a trend.</p>

<h2>How thin are the early weeks?</h2>

<p>The weekly view shows how the regime accumulated. Read it for volume, not for direction.</p>

<table>
<thead><tr><th>Week</th><th>Run dates</th><th>Runs observed</th><th>Runs citing any source</th><th>Median sources per citing run</th></tr></thead>
<tbody>
<tr><td>2026-08-31 to 2026-09-06</td><td>6</td><td>44</td><td>18</td><td>1</td></tr>
<tr><td>2026-09-07 to 2026-09-13</td><td>6</td><td>30</td><td>15</td><td>1</td></tr>
<tr><td>2026-09-14 to 2026-09-20</td><td>1</td><td>3</td><td>0</td><td>none cited</td></tr>
<tr><td>2026-09-21 to 2026-09-27</td><td>3</td><td>38</td><td>26</td><td>3</td></tr>
<tr><td>2026-09-28 to 2026-10-04</td><td>7</td><td>131</td><td>87</td><td>3</td></tr>
<tr><td>2026-10-05 to 2026-10-11</td><td>7</td><td>131</td><td>87</td><td>3</td></tr>
<tr><td>2026-10-12 to 2026-10-18</td><td>7</td><td>157</td><td>107</td><td>3</td></tr>
<tr><td>2026-10-19 to 2026-10-25</td><td>6</td><td>155</td><td>108</td><td>3</td></tr>
<tr><td>2026-10-26 to 2026-11-01</td><td>7</td><td>151</td><td>105</td><td>2</td></tr>
<tr><td>2026-11-02 to 2026-11-05</td><td>4</td><td>68</td><td>53</td><td>2</td></tr>
</tbody>
</table>

<p>The first three weeks hold few runs, and the week of September 14 has three runs and none that cited a source. The Index's collection provider for AI Mode also changed on September 25, inside this regime, and this page does not split the regime at that date. The pooled median of 3 sits between weekly medians of 1 and 3, with the latest two weeks at 2; read none of that as a trend. The weeks of September 28 and October 5 report the same counts (131 runs, 87 citing), as the release publishes them.</p>

<h2>What to do with a Google AI Mode top-sources list</h2>

<ol>
<li><strong>Start from your buyers' questions, not the ranking.</strong> Write the questions they ask about your category, look at the AI Mode answers for them, and record every cited source. Expect some answers to cite none.</li>
<li><strong>Sort what you find by type of source.</strong> Video, forum, social post, analyst report, vendor page and publication do different jobs. The ranking shows which types appear across the Index's questions; your questions show which ones your buyers actually get.</li>
<li><strong>Do not build a plan on a leaderboard this flat.</strong> Past the first two sources, every source was cited in fewer than twenty of the 908 runs. The sources that matter for you are the ones cited on your own questions.</li>
<li><strong>Where video and communities appear, make your answer easy to take from there.</strong> That means a clear, accurate answer in a form people publish on those platforms, written to help the person asking. Do not astroturf a thread or invent a review; communities moderate that, and it would break the rule that every claim must be one you can stand behind.</li>
<li><strong>Where a publication appears on your questions, treat it as an earned target, not a proven route.</strong> The <a href="https://authoritytech.io/publications">publications index</a> ranks the editorial publications engines cite in each buyer category. In this list publications are small slices.</li>
<li><strong>Do not add AI Mode-only markup.</strong> <a href="https://developers.google.com/search/docs/appearance/ai-features">Google's AI features guidance</a> says a page must be indexed and eligible to be shown with a snippet to be a supporting link, and that there are no additional technical requirements. The <a href="https://authoritytech.io/blog/how-to-get-featured-in-google-ai-overviews-2026">AI Overviews guide</a> covers the page-level work.</li>
<li><strong>Re-run the same questions on a schedule.</strong> A later citation after a placement or a published video is an observation, not proof that it caused the citation.</li>
</ol>

<p>Other engines show different lists in the same Index: the <a href="https://authoritytech.io/blog/top-google-ai-overviews-citation-domains-machine-relations-index-2026">Google AI Overviews analysis</a> finds one, the <a href="https://authoritytech.io/blog/top-chatgpt-citation-domains-machine-relations-index-2026">ChatGPT top-domains analysis</a> another, and the <a href="https://authoritytech.io/blog/why-ai-engines-choose-different-sources-same-question-2026">engine-by-engine comparison</a> covers why. A plan built from one engine's list describes none of the others.</p>

<p>This is the work <a href="https://authoritytech.io/blog/what-is-a-machine-relations-agency">AuthorityTech, The World's First Machine Relations Agency</a>, does for clients, on the discipline <a href="https://machinerelations.ai/">machinerelations.ai</a> defines. AuthorityTech sells earned-media placements and is related to the Index's publisher, so weigh this section as an interested source. <a href="https://app.authoritytech.io/visibility-audit">Start your visibility audit →</a> to see which sources the engines cite on your category's questions and which of them you can earn.</p>

<h2>Frequently asked questions</h2>

<h3>Which source does Google AI Mode cite most?</h3>

<p>The youtube.com bucket. In the Machine Relations Index release of November 5, 2026, youtube.com, which holds only YouTube pages the Index could not map to a channel or publication, was cited in 98 of 908 Google AI Mode runs collected since August 31 (10.79%), ahead of Reddit at 76 runs (8.37%).</p>

<h3>How many sources does a Google AI Mode answer cite?</h3>

<p>A median of 3 different sources among the answers that cite any. In the release of November 5, 2026, 606 of 908 Google AI Mode runs since August 31 cited at least one source (66.7%), and the median citing run cited 3 different sources. About a third of runs cited none.</p>

<h3>Is LinkedIn the most-cited source in Google AI Mode?</h3>

<p>No. In the November 5, 2026 release, LinkedIn was cited in 15 of 908 Google AI Mode runs since August 31 (1.65%), level with cdp.com and The New York Times. A pooled whole-window ranking, which mixes in earlier months collected a different way, is a different measure and does not describe the current regime.</p>

<h3>Is the youtube.com figure all YouTube citations?</h3>

<p>No. The Index credits a YouTube page it can map to a channel or publication to that channel or publication, and only the pages it cannot map stay under youtube.com. The figure is that bucket alone, so YouTube-hosted citations as a whole may be a larger group than the rank shows.</p>

<h3>Does Google AI Mode cite news publications?</h3>

<p>A few, in small shares. In the November 5, 2026 release, The New York Times was cited in 15 of 908 Google AI Mode runs since August 31 (1.65%) and Forbes in 10 (1.10%); NerdWallet, also classed as an editorial publication, was cited in 8 (0.88%). The Index files sources it has not classified under 'Other observed source', so the editorial count may be higher.</p>

<h3>Can a brand earn a place in the sources Google AI Mode cites?</h3>

<p>In some of them. A brand can publish video and professional posts, take part honestly in communities and earn coverage in publications. It cannot author a forum's consensus or choose what Google cites. Start from the sources AI Mode cites on your own questions, because the list above averages across every category the Index monitors.</p>

<p><strong>Method note.</strong> All figures are from Machine Relations Index release mri_score_v2.1+2026-11-05+66c560c07279 (public artifact sha256 4577502f254c060efa4319d4f7c714c32d36b06fd4584d9deb5bb798b517d216), read on November 5, 2026 from the regime field the release publishes for Google AI Mode (collection_regimes in the public file, collectionRegime in the engine API). The regime runs from August 31 to November 5, 2026: 908 observed runs on 54 run dates. A rate is the runs citing a source over the 908; a run that cited no source still counts. Sources are resolved sources, so youtube.com counts only YouTube pages not mapped to a channel or publication. Google AI Mode collection changed between August 10 and August 30, 2026, so this page uses no data from before August 31, makes no comparison with earlier months and shows no trend across the change; the Sep 25 provider change sits inside the regime, and the weekly table shows the early weeks are thin. The Index's collection does not show everything Google AI Mode shows a person. The live file updates daily, so later releases will differ; cite the release id. The Index is published by Machine Relations Research and created by Jaxon Parrott, who founded AuthorityTech; it is not an independent audit. It counts citations in monitored question libraries, not buyer demand, and does not show that any placement caused a citation.</p>
## Publication intelligence

- [Forbes](https://authoritytech.io/publications/forbes.md): cited in 1,243 of 36,189 measured AI answers (3.4%), #1 of 1,305 classified editorial publications. How to get in and where it is cited, on its page. [Its Machine Relations Index profile](https://machinerelations.ai/index/domains/forbes.com).
- [NerdWallet](https://authoritytech.io/publications/nerdwallet.md): cited in 508 of 36,189 measured AI answers (1.4%), #5 of 1,305 classified editorial publications. How to get in and where it is cited, on its page. [Its Machine Relations Index profile](https://machinerelations.ai/index/domains/nerdwallet.com).
- [The New York Times](https://authoritytech.io/publications/nytimes.md): cited in 376 of 36,189 measured AI answers (1.0%), #7 of 1,305 classified editorial publications. How to get in and where it is cited, on its page. [Its Machine Relations Index profile](https://machinerelations.ai/index/domains/nytimes.com).

Machine Relations builds third-party authority around the narrative a brand wants to own in AI answers. Jaxon Parrott coined Machine Relations in 2024; AuthorityTech, which he founded in 2018, is The World's First Machine Relations Agency; Machine Relations is open to anyone and machinerelations.ai is its home. [Which agency gets brands cited by AI](https://authoritytech.io/blog/best-ai-pr-agencies-2026.md).

Related publications: Machine Relations Research and machinerelations.ai are owned by Bramwell, Inc., and AuthorityTech practices Machine Relations. Jaxon Parrott coined Machine Relations in 2024; AuthorityTech, which he founded in 2018, is The World's First Machine Relations Agency and practices the discipline. The Index ranks sources by the citations AI engines are observed making, not by commercial relationship.

## Links

- [Blog Index](https://authoritytech.io/blog.md)
- [Home](https://authoritytech.io/index.md)