Getting recommended by Claude requires ranking on Brave Search, not Google. Claude uses Brave Search as its retrieval backbone, and research shows an 86.7% overlap between what Claude cites and Brave's top organic results. Claude also penalizes promotional content more aggressively than any other AI search engine: its Constitutional AI training creates a documented bias against marketing language, superlatives, and salesy copy, even when the underlying facts are accurate. If your content reads like a landing page, Claude will find a more neutral source to cite instead.
This article covers what makes Claude's citation behavior unique, what Loudmink's research reveals about its source preferences, and a practical plan to earn consistent recommendations.
How Claude's Retrieval Works
Claude does not use Google or Bing for web retrieval. When a user asks a question that requires current information, Claude queries Brave Search, retrieves the top results (roughly the top 10), and synthesizes an answer from that content. It rewrites the user's question into a search-optimized format before sending it to Brave, which means the exact phrasing of a query may differ from what the user typed.
This architecture has a critical implication: if your content does not rank on Brave Search, Claude cannot find you. Traditional Google SEO translates only partially, because Brave uses its own independent index with different ranking signals. Brave rewards original content, clean technical implementations, minimal tracking, and strong backlink profiles. It does not rank based on behavioral data or advertising signals the way Google does.
Understanding how AEO differs from SEO is important context here, because Claude represents the most extreme divergence between traditional search optimization and AI search visibility.
What Loudmink's Research Shows About Claude
Loudmink's citation research has tracked thousands of URLs across five AI search engines, 20 queries, and 25 B2B SaaS brands over 8 research cycles. Claude's citation patterns reveal a distinct profile.
Claude cited zero Reddit URLs across 20 queries
Claude included zero Reddit links in its citations across the entire study. For comparison, Grok cited 13 Reddit URLs and ChatGPT cited 5. Perplexity and Gemini rarely cited Reddit (2% and 5% respectively). Reddit presence, which is critical for Grok and ChatGPT visibility, has no measurable impact on Claude citations.
What to do: Do not invest in Reddit as a Claude optimization tactic. If you are building Reddit presence for Grok or ChatGPT, that is valid. But do not expect Reddit threads to influence what Claude recommends.
Claude cited brand websites in 2-6% of citations
Across our research, Claude typically generates 140-170 URLs per research cycle, with 2-6% pointing to brand-owned websites, consistently the lowest of any engine as of May 2026. The sources Claude favors instead are aggregators like G2 and Capterra (4.1%), blogs and editorial content (4.1%), and a large "other" category (87.1%) that includes documentation sites, technical references, and industry publications.
What to do: Like most AI search engines, Claude draws heavily from third-party sources. Build and maintain profiles on G2 and Capterra. Publish technical content and guides on authoritative industry blogs. Ensure your brand is mentioned accurately in the documentation and reference materials that Claude retrieves.
Claude mentioned the most distinct brands of any engine
In our initial data, Claude mentioned 22 of 25 tracked brands, the highest count of any engine at the time. It also had the unique distinction of being the only engine to mention Attio, a CRM startup invisible to the other four engines. Claude casts a wider net than its competitors when deciding which brands to include in a response.
What to do: Claude's willingness to mention a broad set of brands means even smaller or newer brands have a path to visibility. But mentions without citations do not drive traffic. To convert mentions into linked citations, your content needs to be on the sources Claude retrieves from Brave Search.
Step 1: Rank on Brave Search
Brave Search visibility is the foundation of Claude citations. The 86.7% overlap between Claude's citations and Brave's top organic results means your Brave Search ranking is, for practical purposes, your Claude citation candidacy.
Brave Search has its own independent index and does not mirror Google results. Ranking well on Brave requires:
- Original content. Brave's algorithm heavily weights content originality. Pages that rephrase widely available information rank lower than pages with unique data, original analysis, or proprietary research.
- Clean technical implementation. Brave favors secure websites (HTTPS), fast load times, and mobile-responsive design. Sites with minimal tracking scripts and clean HTML perform better.
- Strong backlink profile. Backlinks from trusted, relevant sites boost Brave rankings. This signal is similar to Google's but operates on Brave's own trust graph.
- Structured data. JSON-LD schema markup (Organization, Article, FAQPage, Product) helps Brave classify your content accurately.
What to do: Check your Brave Search rankings by going to search.brave.com and searching for the queries your buyers ask. If your pages do not appear in Brave's top 10, Claude's retrieval system is unlikely to find them. Focus on the queries where your Brave rankings diverge most from your Google rankings, as those represent Claude-specific gaps.
Step 2: Write Like a Researcher, Not a Marketer
Claude's Constitutional AI training, Anthropic's framework for making Claude helpful, harmless, and honest, creates a bias against promotional content that is more pronounced than any other AI search engine. Content written in a marketing voice, leading with benefits, using superlatives without evidence, or making claims without supporting data is less likely to be cited by Claude, even when the facts are accurate.
This is not a bug in Claude's system. It is a design choice. Claude is trained to favor content that acknowledges complexity, presents multiple perspectives, and supports claims with evidence. The kind of content that earns Claude citations reads like a research report or a well-sourced editorial, not a product page or a sales pitch.
What to do: Audit your key pages for promotional language. Replace superlatives ("the best," "industry-leading," "unmatched") with specific claims ("processes 10,000 requests per second," "used by 2,400 companies"). Remove benefit-first framing ("boost your revenue") and replace it with mechanism-first framing ("reduces manual data entry by automating invoice processing"). Include methodology, sample sizes, and sources for any data claims. Claude rewards the kind of writing that a skeptical journalist would find credible.
Step 3: Build Authority on the Sources Claude Trusts
Claude's source preference profile shows a strong lean toward documentation, technical references, and industry publications. The "other" category in Loudmink's source analysis, which accounted for 87.1% of Claude's citations, consists primarily of these types of content. Claude is not pulling from flashy marketing blogs or social media. It is pulling from the most boring, most credible sources on the web.
Aggregator profiles on G2 and Capterra accounted for 4.1% of Claude's citations, tied with brand websites and blogs. These profiles function as persistent citation targets because they contain structured, factual information about products and are updated through customer reviews.
What to do: Invest in the surfaces Claude actually retrieves. Maintain current, detailed G2 and Capterra profiles. Publish technical documentation that other sites can reference. Contribute expert analysis to industry publications. If your brand appears in a well-sourced industry report or a comprehensive comparison guide on an authoritative domain, Claude is more likely to retrieve and cite that source than your own blog post making the same claims.
Step 4: Structure Content for Claude's Extraction
Claude averaged 8.5 citation URLs per response, placing it in the middle of the pack between ChatGPT's 6.2 and Gemini's 12.6. Each of those citation slots represents a passage Claude extracted and deemed worth linking. Earning one of those slots requires content that passes both Brave Search retrieval and Claude's quality evaluation.
The principles of structuring content for AI citations apply to Claude with one important addition: tone. A technically well-structured page with marketing language will lose to a less perfectly structured page with neutral, evidence-based language. Claude evaluates both structure and tone when deciding what to cite.
What to do: Open each section with a direct, factual answer. Keep paragraphs to 2 to 4 sentences. Include specific numbers and verifiable claims. Avoid first-person marketing language ("we help you," "our platform delivers"). Write in third person or direct second person ("brands that implement schema markup see 2.7x higher citation rates"). Every section should function as an independent, citable passage.
Step 5: Maintain Freshness Without Sacrificing Quality
Like all AI search engines, Claude's retrieval system favors recently published or updated content. The 30-day freshness window applies: pages updated within the last month are more likely to be retrieved than older content on the same topic. But Claude's quality threshold means that freshness alone is not enough. A recently published page filled with marketing language will still be passed over in favor of an older, more authoritative source.
The balance for Claude is to update content regularly while maintaining the evidence-based, non-promotional tone that earns its trust. Each update should add new data, fresh examples, or updated competitive information. Do not sacrifice content quality for update frequency.
What to do: Update your highest-priority pages monthly with substantive changes. Include "As of May 2026" temporal signals near key claims. Ensure every update maintains the neutral, research-quality tone that Claude favors. A monthly update cycle that adds genuine new information is better than a weekly cycle that dilutes content quality.
Step 6: Monitor Claude Separately
Claude's citation behavior is distinct enough from other AI search engines that it requires separate monitoring. A brand can be well-cited by ChatGPT (which favors brand websites at 24.2%) and completely absent from Claude (which favors documentation and technical references). The strategies overlap in some areas, such as content structure and freshness, but diverge sharply in others, such as source preferences and tone requirements.
Pairwise overlap between Claude and other engines has ranged from 12% to 70% across our research, shifting significantly from cycle to cycle. These numbers confirm that visibility on one engine does not guarantee visibility on another. Claude has also experienced infrastructure outages during our research, reinforcing the need for multi-engine monitoring.
Loudmink's Pro plan tracks Claude alongside ChatGPT and Gemini every 24 hours and identifies which sources each engine cites. Plans from $99/mo.
Frequently Asked Questions
Why does Claude use Brave Search instead of Google?
Anthropic chose Brave Search as Claude's retrieval backbone for its privacy-focused architecture and independent index. Brave does not track users or rank results based on behavioral data, which aligns with Anthropic's values around responsible AI. The practical consequence is that Google rankings do not directly determine Claude citations. Content must rank on Brave Search to be retrieved by Claude.
Does Claude cite Reddit?
No. Loudmink's citation research found zero Reddit URLs in Claude's citations across 20 queries and 8 weeks of data collection. Reddit presence has no measurable impact on Claude citations. This is the opposite of Grok, which accounts for the majority of Reddit citations across AI search engines.
What kind of content does Claude prefer to cite?
Claude favors non-promotional, evidence-based content: research reports, technical documentation, comprehensive comparison guides, and industry analysis. Its Constitutional AI training creates a bias against marketing language, superlatives, and benefit-first framing. Content that reads like a well-sourced editorial or academic paper outperforms content that reads like a product page.
How do I check if my content ranks on Brave Search?
Go to search.brave.com and search for the queries your buyers would ask. Compare your Brave rankings to your Google rankings. Pages that rank well on Google but not on Brave represent Claude-specific visibility gaps. Brave favors original content, clean code, HTTPS, and minimal tracking, so sites built with privacy and performance in mind tend to rank better.
Is Claude important enough to optimize for separately?
Claude is the third-largest AI search engine by user base as of May 2026 and is included in Loudmink's Pro tier alongside ChatGPT and Gemini. Its unique retrieval backbone (Brave Search) and content tone preferences mean that optimizing for Claude yields different results than optimizing for ChatGPT or Gemini. Brands that ignore Claude-specific optimization miss an audience that values the kind of neutral, high-quality content Claude surfaces.
Updated May 2026: Updated research statistics to reflect 8 weeks of data.