Claude has become one of the most widely used AI assistants for research, comparison, and decision-making queries, and like every AI Search engine it answers by pulling from, and citing, sources across the open web. For brands, that raises a practical question: when Claude composes an answer, which kinds of sources does it actually reach for, and which does it ignore?
OtterlyAI analyzed 379,321 Claude citations across 16,406 domains in June, sorted into 11 source categories. The headline: Claude’s citations are overwhelmingly brand-dominant, spread across a long tail of domains, and almost entirely absent of social media. It behaves very differently from other engines, with direct consequences for how you earn AI citations.
This study looks at what types of sources Claude cites. Specific brand names are intentionally not listed.
Key Findings (TL;DR)
- Reddit is absent from Claude: Social media seems less important for Claude (0.9% of citations) LinkedIn alone is 43.1% of all social citations, while Reddit was cited 0 times
- Brand sites dominate Claude’s citations. Company and product domains account for 64.0% of all citations, more than four times the next category (News and Media at 14.9%). Claude leans on first-party brand content far more heavily than traditional search results do.
- Claude spreads citations across a very long tail. The top 10 domains represent just 9.5% of all citations, and 32.9% of cited domains appear only once. There is no small club of dominant sources.
- Breadth and depth split by source type. Brand wins on breadth (10,842 distinct domains); Social, Encyclopedia, and Video win on depth, a handful of trusted domains cited over and over (96 citations per domain for social).
- Citation behavior doesn’t change often. Across multiple weeks, no source category shifted by more than roughly one percentage point. The mix you optimize for today should still hold next month.
Why Is Studying Claude Critical for AI Search Optimization?
In AI Search, being cited in the answer matters more than ranking, because the answer appears right in the interface and clicks drop.
But “AI Search” is not one channel. When we compared Claude and ChatGPT, only 13% of cited domains overlapped, and around 4.2% at the URL level. ChatGPT leans on community content like Reddit; Claude leans on brand and product pages. Two separate channels, two different sets of winners.
So the real question is which engine your buyers use. Claude skews toward research, comparison, and B2B decisions. If your target users are on Claude, that is where your GEO belongs, focused on the brand-dominant, reference-grade content Claude rewards. Watching ChatGPT alone leaves you blind to Claude.
Want to know if Claude cites your brand?
Claude cites differently from ChatGPT (only 13% of domains overlap), so it needs its own tracking. OtterlyAI now tracks Claude as a dedicated engine: see your mentions, citations, and competitor benchmarks in one dashboard.
👉 Track your brand on Claude with OtterlyAI
Scope of Study
This article summarizes an original OtterlyAI analysis of AI citations generated by Claude for SaaS and technology queries. We tracked, day by day, which domains Claude cited in its answers across a broad, continuously monitored set of SaaS and tech prompts, then classified each cited domain into a source category.
- The AI Search engine (Claude)
- The cited domain
- The domain’s source category (Brand, News and Media, Social Media, Encyclopedia, and so on)
- The date of the citation
- The number of citations recorded for that domain on that date
In total the dataset covers 379,321 citation instances, 16,406 unique domains, 11 source categories, and the entire month of June.
Two interpretive notes:
- This is a descriptive study of citation patterns, not a controlled experiment. It shows what Claude cited, not a proven cause for why any single page was chosen.
- A short window from June 9 to June 13 reflects lighter data collection. We treat the dip in that window as a sampling artifact rather than a real decline, and the category shares below are stable regardless of it.
What Types of Sources Does Claude Cite Most?
Sorted by citation volume, the source categories Claude draws on are strikingly top-heavy. One category, Brand, accounts for nearly two-thirds of everything Claude cites. Everything else competes for the remaining third.
| Source Category | Share | Unique Domains |
|---|---|---|
| Brand | 64.0% | 10,842 |
| News / Media | 14.9% | 1,688 |
| Blogs / Personal Sites | 5.4% | 1,041 |
| Others | 3.7% | 1,162 |
| Education | 2.7% | 518 |
| Government / NGO | 2.4% | 596 |
| Unclassified | 2.3% | 221 |
| Encyclopedia | 2.1% | 142 |
| Community / Forum | 1.3% | 139 |
| Social Media | 0.9% | 37 |
| Video | 0.3% | 20 |
“Unclassified” is mostly country-code domains (for example .co.uk or .com.au) the categorizer could not resolve to a single type.
1. Brand domains dominate, but no single brand owns the space
At 64.0% of all citations, brand domains, meaning company, product, and service websites, are the backbone of Claude’s answers. This is the single most important pattern in the study. When Claude explains a product category, compares options, or answers a “how does X work” question, it overwhelmingly reaches for first-party content published by companies themselves.
Crucially, that dominance is diffuse, not concentrated. The Brand category spans 10,842 distinct domains, and the single most-cited brand domain accounts for under 3% of brand citations. The top 10 brand domains together make up only about 13% of the category. In other words, no brand owns Claude’s attention. The field is wide open and contestable, which is good news for challengers and a warning for incumbents.
What this means for your GEO strategy: your own website is your highest-leverage AI-visibility asset. Treating your product, solution, and comparison pages as authoritative, well-structured reference material, rather than thin marketing copy, is the most direct path to Claude citations.
2. News and Media is a distant but meaningful second
News and Media is the second-largest category at 14.9%, with a higher citation-per-domain rate (34) than Brand. Claude tends to reach for specialist and authoritative publications rather than general-interest headlines, with a recurring presence of financial and business-news outlets in particular. For brands, earned coverage in credible, topically relevant publications remains a real route into Claude’s answers, reinforcing rather than replacing first-party content.
3. Reference sources punch above their weight
Encyclopedia (2.1%), Education (2.7%), and Government and NGO (2.4%) are small by volume but behave like trusted anchors: each is concentrated in a handful of domains cited repeatedly (Encyclopedia averages 56 citations per domain). These reference-grade sources are exactly what an answer engine reaches for when it needs a definitional or authoritative backstop. You will rarely “win” these categories, but understanding them tells you what “citeable” looks like to Claude: clear, structured, factual, and well-sourced.
The Long Tail: Claude Spreads Citations Across Thousands of Sources
If Brand dominance is the first surprise, the second is how widely Claude distributes citations beneath that. This is not a system that recycles the same few authority sites for everything.
| Top N Domains | Citations | Share of All Citations |
|---|---|---|
| Top 10 | 35,924 | 9.5% |
| Top 25 | 68,540 | 18.1% |
| Top 50 | 101,560 | 26.8% |
| Top 100 | 135,899 | 35.8% |
| Top 500 | 227,355 | 59.9% |
| All 16,406 domains | 379,321 | 100% |
The top 10 domains account for just 9.5% of all citations. You need roughly 500 domains to reach 60% of citations, and 32.9% of all cited domains were cited only once. Claude’s sourcing is unusually distributed compared with traditional search, where a small set of high-authority sites tend to dominate.
What the long tail means for GEO
Distribution is opportunity. Because Claude pulls from thousands of sources, you do not need to be a top-10 domain to be cited. A focused page that answers a specific question clearly can earn a citation even if your domain has modest authority. The flip side: breadth of well-structured content matters, because each additional page is another chance to match a query.
The Role of Social Media in Claude’s Citations
This is where Claude diverges most sharply from conventional AI-visibility advice. Social media is not a meaningful citation source for Claude. Across the entire dataset, social platforms accounted for just 0.9% of citations, roughly 3,558 of 379,321. If your AI strategy assumes Claude surfaces social posts, the data says otherwise.
Ranked: social platform share of Claude’s social citations
| Rank | Platform | Share of Social Citations |
|---|---|---|
| 1 | 43.1% | |
| 2 | 14.1% | |
| 3 | TikTok | 9.1% |
| 4 | ResearchGate | 6.7% |
| 5 | 6.7% | |
| 6 | 3.3% | |
| 7 | SlideShare | 3.1% |
| 8 | X (Twitter) | 1.6% |
| – | 0% |
The striking absences: Reddit and X
The most newsworthy finding in this section is what is missing. Reddit, which other published AI-citation research has found to be one of the most-cited social sources on engines such as Perplexity and ChatGPT, was cited zero times by Claude in this dataset. X (Twitter) appeared only 56 times, about 1.6% of an already tiny social slice. The takeaway is blunt: do not port a Reddit-first or X-first AI playbook onto Claude. What works for other engines does not transfer here.
Social citations skew professional, not consumer
What little social content Claude does cite is professional and document-oriented. LinkedIn alone is 43.1% of social citations, and professional or knowledge platforms such as ResearchGate and SlideShare appear well ahead of consumer-first networks like Instagram and TikTok relative to their size. Long-form, profile-and-document content travels into Claude’s answers; ephemeral feed content largely does not.
What this means for your social and GEO strategy
Right-size the effort. For Claude specifically, social media is a low-leverage channel for citations. If you invest at all, prioritize a content-rich LinkedIn presence and authoritative, well-structured documents (the kind that live on professional and knowledge-sharing platforms) over chasing Reddit threads or viral short-form posts. Better still, redirect that energy toward the first-party brand content that makes up 64% of what Claude actually cites.
Claude’s Citation Behavior Is Stable Over Time
A fair question about any monthly snapshot is whether the pattern holds, or whether it is drifting. In this dataset, the category mix is remarkably steady.
Comparing the first five days with the last five, no category moved by more than about one percentage point. Brand held near 64% throughout; News and Media drifted down slightly; social stayed near 1%. Daily citation volume ran in a consistent band of roughly 20,000 to 24,000 per day.
What stability means for planning
Stability is a planning gift. Because Claude’s sourcing behavior is consistent week to week, the optimization choices you make now are unlikely to be invalidated by short-term volatility. You can build a content and GEO roadmap around these patterns with reasonable confidence that they will still describe Claude’s behavior next month.
The Claude Playbook for AI Search
Based on what we observed, here is what we would actually prioritize if your goal is Claude citations and AI visibility.
1. Win with first-party, reference-grade content
Brand domains are 64% of Claude’s citations, and no single brand dominates. Treat your own site as the primary battlefield. Build authoritative product, solution, comparison, and “how it works” pages that read like references, not brochures. Clear definitions, structure, and supporting evidence are what get cited.
2. Structure pages so machines can extract them
The categories Claude cites most heavily per domain (encyclopedic, educational, reference) share a trait: clean structure. Use descriptive headings, concise summaries, well-formed lists and tables, and explicit entities and definitions. Make it easy for an answer engine to lift a specific, self-contained passage.
3. Compete on breadth, not just a flagship page
Because citations spread across a long tail and a third of domains are cited only once, coverage matters. A library of focused pages, each answering a specific question, creates more surface area for matching real prompts than a single comprehensive page can.
4. Right-size your social investment
For Claude, social media is a rounding error, and Reddit and X are effectively absent. If you maintain social for other engines or for audience reasons, fine, but do not expect it to drive Claude citations. Where social does appear, it is professional: LinkedIn and document-style content lead.
5. Build third-party authority signals
News and Media is Claude’s second category, weighted toward credible, specialist publications. Earned coverage in topically relevant, authoritative outlets complements your first-party content and adds independent validation that answer engines value.
Final Conclusion
Our analysis of 379,321 Claude citations points to a clear and actionable picture of how Claude sources its answers:
- Brand-first. First-party company content is 64% of citations, the dominant route to visibility.
- Widely distributed. A long tail of thousands of domains, with the top 10 at under 10%, means modest sites can still be cited.
- Social-light. Social media is 0.9% of citations, with Reddit absent and X negligible. Claude is not a social-driven engine.
- Structure over popularity. Reference-grade, well-structured content is what earns citations, not follower counts or viral reach.
- Stable. The pattern held steady across the study window, so you can plan around it.
If you want visibility inside Claude’s answers, build your website like documentation, not advertising: clear sections, structured data, precise and self-contained passages, and credible sourcing. In AI Search, extractability and reference value determine selection, not channel size or social buzz.
Got a GEO experiment idea? Let’s collaborate
At OtterlyAI, we don’t write about AI Search from the sidelines. We test it. With the community, we run and update a public GEO Experimentation Tracker showing what we are running and what moved AI Search. If you are running your own GEO experiment, email rick.tousseyn@otterly.ai and we may feature it with your name attributed.
👉 Want to measure your AI Visibility? Start with OtterlyAI and our GEO Guide.
👉 Learn how to connect your Claude or ChatGPT to OtterlyAI’s MCP
