By Raul Delapena Setiawan 
The landscape of digital search is undergoing its most disruptive transformation since the inception of the commercial web. As traditional search engine results pages (SERPs) yield ground to artificial intelligence "answer engines," search engine optimization (SEO) professionals are scrambling to decode the black-box algorithms of platforms like OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude.
A series of revelations shared by Jonathan Clark, managing partner at Moving Traffic Media, has shed light on the inner workings of Anthropic’s Claude. Drawing from data presented during a "Zero Click" session by the AI-tracking platform Profound, Clark revealed that Claude’s web search functionality is directly and transparently tied to Brave Search rankings.
Unlike other AI search engines that ingest, synthesize, and dynamically re-rank web results using proprietary algorithms, Claude appears to import and utilize Brave’s top 10 search results directly. This discovery carries profound implications for digital marketers, indicating that Claude may be the most predictable, trackable, and optimizable AI answer engine currently on the market.
Main Facts: The Direct Pipeline Between Claude and Brave Search
The research shared by Clark outlines several defining characteristics of Claude’s search behavior, painting a picture of an AI assistant that behaves quite differently from its primary competitor, ChatGPT.
+------------------------------------------------------------------------+
| CLAUDE VS. CHATGPT: KEY METRICS |
+------------------------------------+-----------------------------------+
| Metric | Claude (Anthropic) | ChatGPT (OpenAI) |
+------------------------------------+-----------------------------------+
| Web Search Frequency | 36.6% of prompts | ~90% of prompts |
| Citation Overlap (Same Prompts) | 8% (Minimal alignment with OpenAI) | 8% (Minimal alignment with Claude)|
| Primary Search Index Partner | Brave Search | Bing / Proprietary Index |
| Result Re-ranking Behavior | Direct trust of top 10 results | Dynamic internal re-ranking |
+------------------------------------+-----------------------------------+1. No Proprietary Re-Ranking
The most significant takeaway from the Profound session is that Claude does not dynamically re-rank the search results it retrieves from the web. Instead, when a query triggers Claude to search the internet, it directly accesses Brave Search’s API and relies heavily on the top 10 results returned by Brave. For SEOs, this means that visibility on Claude is directly correlated with organic performance on Brave Search.
2. Lower Search Frequency Than Competitors
Claude is far more conservative in its use of live web search than ChatGPT. According to the data, Claude executes a web search for only 36.6% of user prompts. In contrast, ChatGPT searches the web for approximately 90% of prompts, representing a massive divergence in how the two systems balance pre-trained parametric memory with real-time retrieval-augmented generation (RAG).
3. Extremely Low Citation Overlap
When presented with identical prompts, Claude and ChatGPT’s citations overlap in only 8% of cases. This minimal overlap confirms that the two platforms rely on fundamentally different indexing pipelines, crawl behaviors, and ranking criteria. While ChatGPT relies heavily on Bing and OpenAI’s emerging proprietary search infrastructure, Claude’s reliance on Brave creates an entirely separate ecosystem of winners and losers in AI visibility.
4. Intent-Specific Search Triggers
Claude’s decision to search the web is highly dependent on the linguistic structure and intent of the user’s prompt. Informational queries seeking static explanations rarely trigger a live search, whereas commercial, transactional, and comparative queries almost always force the LLM to query Brave Search.
Chronology: The Evolution of Anthropic’s Search Capabilities
Understanding how Claude arrived at this architecture requires looking back at the timeline of Anthropic’s development and its strategic partnerships.
+---------------------------------------------------------------------------------------+
| HISTORICAL TIMELINE |
+---------------------------------------------------------------------------------------+
| Early 2021 | Anthropic founded by ex-OpenAI members, focusing on safety/alignment. |
| Mid 2023 | Claude 2 launches; highly capable but entirely offline. |
| Late 2023 | Brave launches its Search API, offering an alternative to Bing/Google. |
| Early 2024 | Claude 3 family debuts. Anthropic quietly integrates Brave's API. |
| Late 2024 | Profound's "Zero Click" session reveals Claude's lack of re-ranking. |
+---------------------------------------------------------------------------------------+- Early 2021: The Foundations of Anthropic. Anthropic was founded by former members of OpenAI, including siblings Daniela and Dario Amodei. From the outset, the company prioritized safety, steering away from real-time web access to avoid the hallucinations, security risks, and prompt injections associated with live web scraping.
- Mid-2023: Claude 2 and the Static Knowledge Limitation. When Claude 2 was released, it was highly praised for its reasoning capabilities and large context window. However, it remained entirely offline, relying on a training cutoff date. Users seeking real-time information, breaking news, or local recommendations had to use alternative platforms.
- Late 2023: Brave Open-Sources and Monetizes its Search API. Realizing the growing demand for independent search indices in the AI era, Brave Software began aggressively marketing its Brave Search API. Positioned as a privacy-centric, independent alternative to Google and Bing, the Brave API was designed specifically to feed LLMs and RAG systems.
- Early to Mid-2024: The Claude 3 Era and Web Integration. With the launch of Claude 3 (Haiku, Sonnet, and Opus) and subsequent iterations like Claude 3.5 Sonnet, Anthropic quietly integrated live web search capabilities. Instead of building an expensive, proprietary web crawler from scratch, Anthropic partnered with Brave to use its independent index.
- Late 2024: The Profound Revelations. During a "Zero Click" session hosted by Profound, data analysts and SEO researchers, including Jonathan Clark, analyzed thousands of queries across ChatGPT and Claude. The findings, published on LinkedIn, confirmed for the first time that Claude does not re-rank Brave’s output, revealing a highly predictable, direct-to-source retrieval model.
Supporting Data: Analyzing Claude’s Query Triggers and Citation Mechanics
To understand how to optimize for Claude, marketers must examine the specific data points regarding when and how the model accesses the web.
When Does Claude Search the Web?
Claude’s search behavior is highly polarized based on the nature of the prompt. It relies on its internal training data for educational or systemic questions, reserving live search for time-sensitive, commercial, or comparative queries.
+-----------------------------------------------------------------------------------+
| CLAUDE WEB SEARCH TRIGGERS |
+-----------------------------------------+-----------------------------------------+
| Triggers Search (RAG Active) | Avoids Search (Uses Internal Memory) |
+-----------------------------------------+-----------------------------------------+
| • "Best [product/service]" | • "How does [process/system] work" |
| • "Top [companies/tools] in [year]" | • "What is the definition of [term]" |
| • "[Brand A] vs [Brand B]" | • "Steps to perform [basic task]" |
| • "[Service] near me" | • Creative writing & coding prompts |
+-----------------------------------------+-----------------------------------------+According to the data shared by Clark, informational prompts containing phrases like "how does," "what is," and "steps to" are highly unlikely to trigger a web search. For these queries, Claude draws entirely on its pre-existing weights. Because no web search is executed, Claude cannot cite external sources for these queries, presenting a zero-traffic-referral scenario for publishers.
Conversely, queries containing terms like "best," "top," "near me," or direct product comparisons consistently trigger Claude to query Brave Search. Once triggered, Claude retrieves the top 10 organic results from Brave, synthesizes the information, and presents the answer with direct citations back to those top-ranking pages.
The Mechanics of the 8% Citation Overlap
The remarkably low 8% citation overlap between Claude and ChatGPT highlights the divergence of the AI search ecosystem:
$$textCitation Overlap = fractextShared CitationstextTotal Citations Generated approx 8%$$
This disparity stems from three primary factors:
- Index Differences: ChatGPT utilizes Bing’s index, which is heavily influenced by Microsoft’s proprietary ranking algorithms, legacy domain authority, and massive web-crawling infrastructure. Claude utilizes Brave Search, which uses an independent index built on different crawl priorities and user-driven discovery mechanisms.
- Algorithmic Filtering: ChatGPT applies secondary, proprietary filtering layers to re-rank search results based on internal quality scores, brand safety, and alignment. Claude, as Clark noted, bypasses this step, accepting Brave’s top results as-is.
- Prompt Interpretation: Even when presented with the same prompt, the internal reasoning engines of GPT-4o and Claude 3.5 interpret user intent differently, leading to slightly different API calls to their respective search backends.
Official Responses and Industry Commentary
While Anthropic has historically kept the exact engineering details of its search integration close to the chest, the company’s documentation and public statements align with these findings. Anthropic has consistently emphasized "constitutional AI" and algorithmic efficiency. Bypassing the computational expense of an internal re-ranking model for search results fits perfectly with Anthropic’s goal of keeping inference costs low and response times fast.
Brave Software, on the other hand, has been highly vocal about its role in powering the AI revolution. In press releases regarding its Search API, Brave has explicitly stated that its index is optimized for AI training and RAG applications, boasting high-speed endpoints and a lack of bias compared to traditional big-tech search engines.
Industry experts have reacted to Clark’s shared findings with enthusiasm, viewing it as a rare window of clarity in an otherwise opaque AI landscape.
"For years, SEOs have lamented that AI engines are a black box," said Jonathan Clark in his analysis. "But if Claude is directly serving Brave’s top results without an internal re-ranking layer, then we finally have a clear, actionable target. If you rank on Brave, you rank on Claude."
Implications: The Rise of Generative Engine Optimization (GEO)
The direct correlation between Brave Search and Claude’s outputs has profound implications for digital marketers, SEO specialists, and brand managers. It shifts the conversation from theoretical "AI optimization" to practical, actionable search tactics.
1. Brave Search is No Longer a Niche Concern
Historically, search marketers ignored Brave Search due to its low market share compared to Google. However, because Brave serves as the data engine for Claude—one of the fastest-growing AI platforms in the world—optimizing for Brave’s index has suddenly become a high-priority channel.
To rank on Brave, and subsequently win citations in Claude, marketers must focus on:
- Direct Submission and Indexing: Ensuring sites are fully crawled by Brave’s independent crawler, Bravebot. Marketers should use Brave’s webmaster tools to monitor indexing status.
- On-Page SEO Fundamentals: Because Brave’s algorithm is less reliant on the complex, user-behavior-driven signals that Google uses (such as Chrome data), it places a higher premium on traditional on-page factors: clean HTML structure, keyword-optimized H1/H2 tags, descriptive alt text, and fast page load speeds.
- Schema Markup: Implementing robust schema markup (Product, Organization, LocalBusiness) helps Brave’s API parse page content cleanly, making it easier for Claude to extract and cite specific details like pricing, ratings, and locations.
2. The Fragmentation of SEO
The low citation overlap between Claude and ChatGPT indicates that search engine optimization is fragmenting into distinct sub-disciplines. Marketers can no longer rely on a single "Google-first" strategy to win across all platforms.
+---------------------------------------------------------------------------------------+
| THE FRAGMENTED SEARCH LANDSCAPE |
+-------------------+----------------------------------+--------------------------------+
| Platform | Primary Search Engine Partner | Optimization Focus |
+-------------------+----------------------------------+--------------------------------+
| Gemini | Google | SGE, EEAT, Chrome User Signals |
| ChatGPT | Bing / OpenAI Search | Bing Webmaster, Brand Mentions |
| Claude | Brave Search | On-Page SEO, Bravebot Indexing |
+-------------------+----------------------------------+--------------------------------+A brand that dominates Google may find itself completely invisible on Claude if it neglects its presence in Brave’s index, or invisible on ChatGPT if Bing has de-indexed its domain.
3. The "Zero-Click" Threat and the Value of Commercial Queries
Because Claude only searches the web for ~36.6% of queries, informational content (e.g., "how-to" guides, definitions) is highly susceptible to traffic loss. Claude will answer these questions directly using its internal memory, leaving no opportunity for organic click-throughs.
Consequently, content strategies must pivot. Instead of targeting broad informational keywords, publishers and brands should focus on creating high-value, middle- and bottom-of-funnel content. Queries containing terms like "best," "vs," and "near me" are highly likely to trigger Claude’s search mechanism, presenting the highest opportunity for generating high-intent organic referral traffic via citations.
4. Continuous Testing and Monitoring
Clark’s observation that years and dates frequently appear in Claude’s search behavior suggests that the engine is highly sensitive to freshness. To maintain visibility in Claude, content must be updated regularly with current year signifiers (e.g., "Best Project Management Software [Current Year]"). Marketers should establish continuous testing protocols, querying Claude directly with key commercial phrases to monitor whether their brands are cited in the generated answers.
Ultimately, the revelation of the Claude-Brave connection simplifies what had previously been a highly complex optimization challenge. By mastering the fundamentals of Brave Search, brands can systematically secure their visibility within one of the world’s most sophisticated artificial intelligence engines.
Laily UPN 
Lina Irawan 
Asep Darmawan