Nobody has published a large-sample ranking of the domains Claude cites. Ahrefs, Semrush, and Profound do not have one in public. The measured fact is the mix. In Otterly's June 2026 file, 379,321 citations: brand domains 64.0%, news and media 14.9%, encyclopedia 2.1%, community 1.3%, video 0.3%. Name lists that circulate without shares are not used as a leaderboard on this page.
- There is no Claude top 10 of domains with shares. Ahrefs' public series skips Claude. Its Brand Radar API, docs updated 18 September 2026, supports Claude only for custom prompts.
- Brand domains are the majority in both opened category studies: Otterly 64.0% in June 2026, Profound 66% in a January 2026 table whose sample sentence does not match the engines it displays.
- News is the next block, 14.9% at Otterly and 24% media at Profound. Encyclopedia and social are small. Video is 0.3% of Otterly's citations, across 20 domains.
- The top 10 domains, whoever they are, were 9.5% of Otterly's citations. The top 100 were 35.8%. 32.9% of cited domains appeared once. The names of that top 10 were not in the text retrieved.
- Otterly's page returned 403 on a direct fetch. The category figures are from the indexed article. A hidden table of domain names cannot be ruled out, and it was not read.
The list that has not been published
This is the page people expect to be a top ten. Publishing invented names would be worse than the gap. Broken chain on any Claude domain rank that does not show a sample, a date, and a denominator. Syntheses that name Wikipedia, Reddit, and a handful of magazines as Claude's top five, without shares and without a Claude sample size, were opened and set aside. They conflict with the category studies below.
Semrush's 13-week study of more than 230,000 prompts did not include Claude. Ahrefs' June 2025 assistant cut did not include Claude. The monthly "50 most-cited websites" series, updated 2 September 2026, covers ChatGPT, Gemini, Perplexity, Copilot, Grok, and AI Mode. The related-post list on the ChatGPT article has no Claude installment.
The mix that has
Otterly, published 30 June 2026, is the cleanest Claude-only category file. 379,321 citations, 16,406 domains, June.
| Source type | Share of citations | Domains in the class |
|---|---|---|
| Brand | 64.0% | 10,842 |
| News / media | 14.9% | 1,688 |
| Blogs / personal | 5.4% | 1,041 |
| Education | 2.7% | 518 |
| Government / NGO | 2.4% | 596 |
| Encyclopedia | 2.1% | 142 |
| Community / forum | 1.3% | 139 |
| Social media | 0.9% | 37 |
| Video | 0.3% | 20 |
Within the social slice only, LinkedIn was 43.1%, Facebook 14.1%, TikTok 9.1%, and Reddit 0%. That is a share of a 0.9% slice. It is not evidence that Claude never cites Reddit on a general-knowledge question. It is evidence that, in this file, Reddit was not where the social citations went.
Profound's 8 January 2026 category table has a Claude row: brand 66%, media 24%, institution 7%, social 3%. Institution includes government, universities, and Wikipedia, so the 7% is not a government share and not a Wikipedia share. The article's opening sample is 27 million citations across ChatGPT, Gemini, and AI Overviews, while the table also includes Claude, Perplexity, AI Mode, and Copilot. The Claude row's own citation count is not stated. Use it as a second look at the mix. Do not add it to Otterly.
A later Profound language study, March 2026 data, published 21 April, puts Claude's social citation rate at 3.99% across a 3.25 billion citation set. That matches the direction of the small social slice. It is still not a domain list.
A 75-prompt name list, labelled as one
Foglift collected 75 brand-neutral buyer-intent prompts on 18 May 2026, five engines, one response each. Claude's top five by distinct prompts, not by share of Claude citations: healthline.com on 6 prompts, fortune.com on 5, consumerreports.org on 4, bettertrail.com on 3, chewy.com on 3. The authors say the lower half of their overall top 100 is volatile, and that informational prompts would tilt toward institutional and encyclopedic sources.
Those names are real. They are not a general Claude ranking. A buyer-intent set is exactly the set that under-represents Wikipedia, which is why it must not be used to "fill in" the missing leaderboard. The same warning applies to SoRank's 0.03% Wikipedia answer rate: 20,011 Claude answers, 79% French, buyer questions, and the authors say to read it as direction.
When Claude cites at all
Anthropic's product post of 20 March 2025 says that when Claude incorporates information from the web, it provides direct citations. Web search was global on all plans from 27 May 2025. The API docs, opened 3 October 2026, say citations are always enabled for the web-search tool, and that Claude searches when the request depends on current or specific facts. It answers without searching for established facts, math, coding concepts, and analysis of material already in the conversation.
Document citations are a second feature. The caller passes a PDF or text with citations turned on, and Claude cites locations inside that file. Counting an uploaded PDF as a web domain would invent a source. Tracker "Claude citation share" is a sample of answers where retrieval ran. It is not a sample of all Claude answers. The rest of that product behaviour is on Claude citation statistics.
What this does not prove
Open question It does not name the ten domains. Otterly says the top 10 were 9.5% of citations and does not, in the text retrieved here, say which ten. It does not say which search index Claude uses. Anthropic's web-search docs do not name a provider. A claimed 86.7% overlap with Brave, attributed to Profound through a secondary audit, was not opened at the original and is not used.
It also does not say Claude ignores reference works in uncited answers. A model can lean on an encyclopedia in training and rarely footnote it. The reference page keeps Wikipedia's measured shares, engine by engine, including this small Claude slice: most-cited reference domains.
There is no public large-sample list of domains Claude cites. The measured mix is about 64% brand domains and about 15% news.
Share on XVerification status
Category shares are Otterly's and Profound's, with the sample mismatch on Profound stated. Foglift's names are prompt counts inside 75 commercial prompts. Synthesised top fives without shares were excluded. The Otterly URL was not fully fetched. No domain was added to make the table look finished.
Namdev, R. (2026). Most-cited domains in Claude (v1). Retrieved from https://ritiknamdev.com/blog/most-cited-domains-claude Published under CC BY 4.0 — reuse freely with attribution.
What is known about Claude's citations beyond this gap is on Claude citation statistics.