April 7, 202616 min read

Claude (Anthropic): How Claude Recommends Brands and How to Optimize for It

Deep dive into Claude by Anthropic: how it forms brand recommendations, what data sources it uses, how it differs from ChatGPT, and how to increase your brand visibility in Claude answers.

Vladislav Puchkov

Founder of GEO Scout, GEO optimization expert

Claude by Anthropic occupies a unique position among AI providers. It is not the most popular (that is ChatGPT), not the most comprehensive in coverage (that is Gemini), not the most generous with recommendations (that is Grok). But Claude is the most careful. The most thoughtful. And, importantly for brands — the most citation-heavy.

According to GEO Scout monitoring data from March 2026, Claude shows a distinctive pattern: it either knows a brand at 80-100%, or does not know it at all (0-10%). There are few intermediate values. At the same time, its domain citation — the percentage of answers where Claude places a direct link to the brand's domain — is among the highest across all providers.

This article is a deep dive into how exactly Claude forms brand recommendations, how it differs from ChatGPT, and how to optimize your brand for visibility in Claude's answers.

What is Claude (Anthropic)

Company History

Anthropic was founded in 2021 by Dario and Daniela Amodei — former OpenAI researchers who left the company over disagreements about AI safety. Anthropic's mission is building safe, reliable, and interpretable AI systems. The company quickly became one of the three leading AI startups in the world alongside OpenAI and Google DeepMind.

Claude Models: Evolution

In February 2026, Anthropic released Claude Opus 4.6 with a 1 million token context window — roughly 750,000 words. The model scores 78.3% on the MRCR v2 benchmark at maximum context. Claude Sonnet 4.6 became the default model for free and paid plans — users preferred it over the previous flagship Opus 4.5 in 59% of cases.

Claude's Positioning

Claude is positioned as a "thoughtful" AI assistant. Unlike ChatGPT, which aims to give a quick answer, Claude more often clarifies context, offers multiple options, and honestly states its limitations. This is not a bug — it is an architectural decision embedded in Constitutional AI.

How Claude Forms Answers About Brands

Constitutional AI — the Foundation of the Approach

Constitutional AI is a method developed by Anthropic for training models. Instead of relying solely on RLHF (reinforcement learning from human feedback), Anthropic uses RLAIF — reinforcement learning from AI feedback. The model critiques and corrects its own responses based on a "constitution" — a set of principles.

In January 2026, Anthropic published Claude's new constitution — a comprehensive 80-page document. The core priorities, in descending order:

1. Broadly safe — not undermining appropriate human oversight of AI
2. Broadly ethical — being honest, acting according to good values
3. Compliant with Anthropic's guidelines — following specific company instructions
4. Genuinely helpful — benefiting operators and users

For brand recommendations, this has direct consequences. Claude:

• Avoids unjustified preferences. If the model lacks confident data about a brand's superiority, it will present multiple options instead of one
• Warns about limitations. Unlike ChatGPT, Claude more often says "I don't have current data about..." or "the recommendation depends on your specific needs"
• Strives for honesty. If a brand has mixed reviews, Claude will mention both positive and negative aspects
• Values sources. High domain citation is not accidental — it is a consequence of the architectural decision to rely on verifiable information

Claude's Three Data Sources

Claude forms answers from three sources, each managed by a separate crawler.

1. Training Data (ClaudeBot)

ClaudeBot is the crawler that collects public web content for training Anthropic's models. It works similarly to OpenAI's GPTBot: crawls the internet, collects content, includes it in the training corpus.

Characteristics:

• Content collected by ClaudeBot forms the model's "parametric memory" — knowledge embedded in neural network weights
• Blocking ClaudeBot excludes future content from training data but does not remove already-collected content
• The old user-agent strings Claude-Web and Anthropic-AI are deprecated and no longer function

For brands: if ClaudeBot does not scan your site, future Claude versions will know less about you. This is especially critical for niche and new brands.

2. Real-Time Web Search (Claude-User)

Claude-User activates when a user asks a question that requires current information from the internet. This is not a scheduled crawl — it is a real-time fetch.

When it triggers:

For brands: if Claude-User is blocked in your robots.txt, Claude cannot load your pages when answering user questions. Your competitors will answer — you will not.

3. Search Index (Claude-SearchBot)

Claude-SearchBot is the newest component of Anthropic's infrastructure. It indexes content for Claude's search feature, which received a major update in February 2026: Dynamic Filtering with source citations.

Blocking Claude-SearchBot is equivalent to blocking Bingbot for Bing. Your site simply will not appear in Claude's search results.

robots.txt Configuration for Claude

Anthropic is the only major AI provider that guarantees all three crawlers respect robots.txt, including Claude-User. OpenAI, by comparison, explicitly states that ChatGPT-User "may not follow" robots.txt.

Recommended configuration for brands:

# Allow all three Claude crawlers (default)
 
# If you want to block training only:
User-agent: ClaudeBot
Disallow: /
 
# Claude-User and Claude-SearchBot are allowed by default

Check for deprecated entries! If your robots.txt contains Claude-Web or Anthropic-AI — they no longer work. Replace them with ClaudeBot, Claude-User, and Claude-SearchBot.

Claude vs ChatGPT: Differences in Brand Recommendations

GEO Scout data from March 2026 shows a systematic difference between how Claude and ChatGPT recommend brands.

Recommendation Approach

Specific Data: Claude vs ChatGPT by Niche

EdTech:

Claude knows Russian EdTech significantly better than ChatGPT. SkillFactory — 72.4% vs 6.7%. An 11x gap. However, ChatGPT knows international brands better.

Hosting:

Claude confidently leads in knowledge of Russian hosting providers. AdminVPS — 62.9% at Claude and 0% at ChatGPT. Timeweb — 94.3% vs 43.3%.

FinTech:

Here the "all or nothing" pattern manifests: MTS Bank — 0% at Claude, while VTB and Alfa get 63.3%.

E-commerce:

The only niche where ChatGPT confidently wins. Claude is less certain about marketplaces — 70% vs 100%.

Domain Citation: Claude's Unique Advantage

Claude places direct links to brand domains more often than other providers. This is not accidental — it is a consequence of the Constitutional AI architecture that values verifiable information.

62.1% domain citation for Yandex Travel means: in 62 out of 100 answers, Claude places a direct link to travel.yandex.ru. That is direct AI traffic — without SEO, without contextual ads.

Which Brands Claude Recommends Most

Based on GEO Scout data, we identified three key patterns.

Pattern 1: "All or Nothing"

Claude either knows a brand at 80-100%, or does not know it at all (0-10%). There are few intermediate values. This distinguishes it from Gemini, which gives even coverage, and from ChatGPT, which more often gives middle values.

Examples of "all": Skillbox (100%), Timeweb (94.3%), Beget (71.4%), Netology (86.2%)

Examples of "nothing": MTS Bank (0%), Skypro (low), many international EdTech brands

Pattern 2: Expert Content = High Positions

Brands with strong expert content (reviews, comparisons, guides) receive from Claude not only mentions but also high positions and domain citation. This is not correlation — it is causation. Claude values content that helps the user make a decision.

Pattern 3: Better Knowledge of Niche Local Brands

Claude knows Russian niche brands better than ChatGPT: hosting (AdminVPS 62.9% vs 0%), EdTech (SkillFactory 72.4% vs 6.7%). At the same time, it knows large international e-commerce less well (Ozon 70% vs 100% at ChatGPT).

How to Optimize Your Brand for Claude

Technical Optimization

1. Configure robots.txt for All Three Crawlers

This is the fastest and technically simplest step — and one of the most important.

# Recommended configuration for brands
User-agent: ClaudeBot
Allow: /
 
User-agent: Claude-User
Allow: /
 
User-agent: Claude-SearchBot
Allow: /
 
# If you want to limit crawl speed:
User-agent: ClaudeBot
Crawl-delay: 1

Common mistake: blocking all AI bots with a single rule in robots.txt. As a result, not only training but also search gets blocked. According to BuzzStream, 71% of top 100 news sites accidentally block AI search bots while trying to block only training.

2. Check for Deprecated User-Agent Strings

If your robots.txt contains Claude-Web or Anthropic-AI — they do not work. Replace with current ones: ClaudeBot, Claude-User, Claude-SearchBot.

3. Add llms.txt

llms.txt is a proposed standard for communicating with AI systems. It complements robots.txt: where robots.txt tells crawlers what they can access, llms.txt provides a structured summary of your site.

# llms.txt — example for a brand
# Name: [Company Name]
# Description: [Brief description]
# Key pages:
# - /about — About the company
# - /services — Services
# - /pricing — Prices
# - /blog — Blog with expert content

Learn more about llms.txt in our article What is llms.txt.

4. Structured Data (Schema.org)

Claude, like other AI providers, uses structured data to extract facts. Key markup types for brands:

• Organization — basic company information
• Product / Service — products and services
• FAQ — answers to frequently asked questions
• Review / AggregateRating — reviews and ratings
• HowTo — instructions and guides

Learn more in our article FAQ Schema markup for AI answers.

Content Claude Prefers

Claude values content that helps the user make a decision. Not promotional — but experimental, factual.

Content Types with Maximum Citability:

Content Format for Claude

Claude extracts information better from content that:

1. Starts with a direct answer. Not "In the modern world of digital technologies..." but "Timeweb is a Russian hosting provider with data centers in Moscow and St. Petersburg."
2. Uses lists and tables. Claude extracts facts from tables more efficiently than from continuous text
3. Contains specific numbers. "10,000 clients" is better than "many clients"
4. Provides context. "Best for small business" is better than "best"

E-E-A-T Signals Important for Claude

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is a concept originating from SEO but gaining special significance for Claude. Constitutional AI directly instructs the model to value trustworthy sources.

Experience

• Publish content from named real experts with biographies
• Add publication and update dates
• Show practical experience using the product

Expertise

• Create deep content with unique data
• Publish research based on proprietary data
• Participate in expert discussions (hubs, forums, media)

Authoritativeness

• Mentions in authoritative sources: media, industry rankings
• Publications on third-party platforms (guest posts, interviews)
• Active presence in professional communities

Trustworthiness

• Accurate contact information on the site
• Clear privacy policy
• Real customer reviews (with verification)
• Updated content — Claude considers data freshness

Content Strategy for Claude: Checklist

1. Homepage: clear company description, products, advantages, specific numbers
2. About page: history, team, expertise, achievements — how to optimize
3. FAQ section: 10-20 questions with specific answers, marked up with Schema.org
4. Blog: regular expert articles with data, comparisons, and case studies
5. Product/service pages: detailed descriptions with comparison tables
6. Reviews and cases: real customer stories with results
7. Pricing page: transparent plans with specific numbers — how to prepare

Practical Steps: A 30-Day Action Plan

Week 1: Diagnostics

Day 1-2: Audit robots.txt

Check your robots.txt for:

• Deprecated user-agent strings (Claude-Web, Anthropic-AI)
• Accidental blocks of Claude-User or Claude-SearchBot
• Absence of entries for Claude bots (which means they are allowed by default)

Day 3-5: Check Visibility in Claude

Sign up at GEO Scout and check:

• Your brand's mention rate in Claude
• Comparison with competitors
• Domain citation
• Positions in answers

Day 6-7: Competitor Analysis

Find 3-5 competitors that Claude recommends more often. Analyze:

• Which pages on their sites rank better
• What content they publish
• What structured data they use

Week 2: Technical Optimization

Day 8-10: Configure robots.txt

Add explicit entries for Claude bots. Ensure Claude-SearchBot and Claude-User are allowed.

Day 11-14: Structured Data

Add Schema.org markup:

• Organization on the homepage
• FAQ on question pages
• Product / Service on respective pages
• Review / AggregateRating for reviews

Weeks 3-4: Content Strategy

Day 15-21: Expert Content

Create 2-3 materials that Claude tends to cite:

• Comparative review with a table
• FAQ with 10-15 questions
• Research with unique data

Day 22-30: External Presence

• Publish a guest post in industry media
• Update brand pages in directories and rankings
• Add or update a Wikipedia article (if applicable)
• Create English-language versions of key pages

FAQ

How does Claude form answers about brands?

Claude uses two sources: training data (information collected by ClaudeBot when crawling the web) and real-time web search (via Claude-User and Claude-SearchBot). In January 2026, Anthropic published a new "Claude constitution" — an 80-page document defining the model's values. Claude strives to give balanced, honest answers without unjustified recommendations, making it more cautious compared to ChatGPT and Gemini.

What are ClaudeBot, Claude-User, and Claude-SearchBot?

These are three separate Anthropic crawlers. ClaudeBot collects data for training models. Claude-User fetches pages in real time when a user asks a question. Claude-SearchBot indexes content for Claude's search feature. Each is controlled separately via robots.txt. Blocking ClaudeBot does not affect Claude-SearchBot and vice versa.

Should I allow ClaudeBot in robots.txt?

For commercial brands — yes. Blocking ClaudeBot means your content won't be included in training data for future Claude versions, and the model will "know" less about your brand. However, you can block ClaudeBot while allowing Claude-SearchBot and Claude-User — this way you protect content from being used in training but remain visible in Claude search.

Why does Claude recommend brands differently than ChatGPT?

Claude uses Constitutional AI — an approach where the model evaluates its answers against a set of principles (honesty, safety, helpfulness). This makes Claude more cautious: it gives direct recommendations less often, presents multiple options more frequently, and warns about limitations. ChatGPT is more inclined to give unambiguous single-brand recommendations.

Which brands does Claude recommend most often?

According to GEO Scout monitoring data, Claude shows an "all or nothing" pattern: it either knows a brand excellently (Skillbox — 100% mention rate, Timeweb — 94.3%) or doesn't know it at all (MTS Bank — 0%). Claude recommends brands with strong expert content and high domain citation more often — for example, Yandex Travel's domain citation reaches 62.1%.

How do I optimize my brand for Claude?

Key steps: allow ClaudeBot, Claude-User, and Claude-SearchBot in robots.txt; create expert content with specific facts and figures; use Schema.org markup; publish reviews and comparisons; ensure English-language presence. Claude values E-E-A-T signals — expertise, experience, authoritativeness, and trustworthiness. Monitor results through GEO Scout (geoscout.pro).

What is Constitutional AI and how does it affect brand recommendations?

Constitutional AI is a training method developed by Anthropic where the model evaluates its answers against a "constitution" — a set of values and principles. In January 2026, Anthropic published an updated 80-page constitution. For brand recommendations, this means: Claude strives for honesty, avoids unjustified preferences, presents multiple options instead of one, and directly states when it has insufficient data.