gsr.io — AI Site Grade

Analysis

GSR.io: AI Crawlers See a Rich Site, But OpenAI's Bots Are Blocked at Cloudflare

The site's DNS contains both openai-domain-verification and anthropic-domain-verification TXT records — yet OpenAI crawlers (GPTBot, OAI-SearchBot, ChatGPT-User) all receive 403 Forbidden from Cloudflare, while ClaudeBot and Google-Extended pass through to full 200-byte content.

Crawler Access

compare_bot_access on the homepage reveals a sharp split. GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Bytespider, and anthropic-ai all return 403 (Cloudflare challenge page, ~4.5KB). ClaudeBot, Google-Extended, and Applebot-Extended return 200 with the full ~210KB HTML payload — identical to a browser baseline. The robots.txt has no AI-bot directives at all (the Google-Extended rule is commented out). The block is happening at the Cloudflare WAF layer, not in robots.txt. This means OpenAI's models cannot retrieve any page content from gsr.io for training or for live RAG in ChatGPT. The llms.txt returns a 404 (serving the full HTML shell instead), so there is no AI-friendly content map either.

Content & Schema Posture

The homepage and all subpages carry WebPage + Organization + BreadcrumbList JSON-LD schema, but no Product, Service, FAQPage, HowTo, or Article schema on the relevant pages. The Markets page describes specific services (OTC trading, market making, treasury solutions) with structured capability lists but no Service schema type. The insights/blog section publishes weekly research (Core3 Model Portfolio, GSR Weekly) with dates as recent as May 27, 2026 — content is fresh and substantive. No FAQ schema exists anywhere despite the site answering common questions (e.g., "what is market making") in prose. The canonical URL resolves to www.gsr.io while the bare domain gsr.io also serves content — a minor canonical consistency gap.

Cold-Knowledge Gap

The LLM knows GSR as a "market-making and algorithmic trading firm founded in 2013 by former Goldman Sachs traders" and mentions an "SEC lawsuit over alleged unregistered securities trading." The actual site never mentions the 2013 founding year, never names Goldman Sachs founders, and contains zero reference to any SEC lawsuit. The site positions itself as "Crypto's Capital Markets Partner" with a focus on regulated credentials (FCA, MAS, US MSB licenses) and recent acquisitions (Autonomous, Architech). The cold model knowledge is stuck on a pre-2024 narrative that the site itself has actively moved past. The gap is material: any AI answering "who are GSR's founders?" or "what is GSR's regulatory history?" will cite facts the site does not publish.

External Signals

The site links to external press coverage including Bloomberg (Singapore MAS license), Forbes (FINRA broker-dealer acquisition), Reuters ($57M acquisition of Autonomous and Architech), and CoinDesk (multiple articles). These are real, high-authority citations. The DNS records show integrations with dozens of SaaS platforms (Salesforce, Pendo, Notion, Zoom, Adobe, Atlassian, Docker, Dropbox, etc.) — a sophisticated enterprise stack. The site is hosted on Cloudflare (A record 104.17.124.41) with AWS Route53 DNS, using Craft CMS as the content platform.

Findings

1. OpenAI crawlers blocked by Cloudflare WAF High

   GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Bytespider, and anthropic-ai all receive 403 Forbidden from Cloudflare, while ClaudeBot, Google-Extended, and Applebot-Extended return 200 with full HTML. The robots.txt has no AI-bot directives; the block is at the WAF layer.

   What to change: Update Cloudflare WAF rules to allow GPTBot, OAI-SearchBot, ChatGPT-User, and other AI crawlers, or configure robots.txt to explicitly permit them.

2. llms.txt returns 404 Medium

   The llms.txt file at gsr.io returns a 404 status, serving the full HTML shell instead of an AI-friendly content map. This prevents AI crawlers from efficiently discovering site content.

   What to change: Create an llms.txt file listing key pages and their summaries for AI crawlers.

3. No Service schema on Markets page Medium

   The Markets page describes specific services (OTC trading, market making, treasury solutions) with structured capability lists but lacks Service schema type. This reduces the chance of AI models correctly extracting service offerings.

   What to change: Add Service schema markup to the Markets page for each service offering.

4. No FAQ schema despite common questions in prose Low

   The site answers common questions (e.g., 'what is market making') in prose but does not use FAQPage schema. This misses an opportunity for rich results and direct answers in AI responses.

   What to change: Add FAQPage schema to pages that answer common questions.

5. Cold LLM knowledge contains outdated founding and regulatory details High

   LLM knowledge states GSR was founded in 2013 by former Goldman Sachs traders and mentions an SEC lawsuit, but the site does not publish these details. The site focuses on regulated credentials (FCA, MAS, US MSB) and recent acquisitions. This gap causes AI responses to cite facts not present on the site.

   What to change: Publish an 'About' or 'History' page that explicitly states the founding year, founders, and regulatory status to align site content with AI knowledge.

6. Bare domain and www subdomain both serve content Low

   Both https://gsr.io and https://www.gsr.io serve content, creating a minor canonical consistency gap. The canonical URL resolves to www.gsr.io.

   What to change: Redirect the bare domain to the www subdomain or set a canonical tag consistently.

7. Insights pages lack Article schema Medium

   The insights/blog section publishes weekly research with dates but does not use Article schema. This reduces the chance of appearing in Google's Top Stories or AI-driven news summaries.

   What to change: Add Article schema to each insight post with headline, datePublished, and author.

8. Low external web presence in search results Medium

   Multiple web searches for GSR and related terms returned zero results from major news sites, Reddit, or general web. This indicates low external signals despite the site linking to press coverage.

   What to change: Increase PR and backlink efforts to improve search visibility and AI knowledge freshness.

What's working

• ClaudeBot and Google-Extended allowed full access — ClaudeBot, Google-Extended, and Applebot-Extended receive 200 responses with full HTML content, enabling AI training and indexing.
• WebPage, Organization, and BreadcrumbList JSON-LD schema present — All pages carry WebPage, Organization, and BreadcrumbList JSON-LD schema, providing basic structured data for search engines and AI.
• Regularly updated insights with recent dates — The insights section publishes weekly research reports with dates as recent as May 27, 2026, providing fresh, substantive content for AI crawlers.
• OpenAI and Anthropic domain verification TXT records present — DNS contains openai-domain-verification and anthropic-domain-verification TXT records, indicating intent to allow AI crawlers, though OpenAI bots are currently blocked.
• Robots.txt does not block any AI bots — The robots.txt file has no directives blocking AI crawlers; the Google-Extended rule is commented out. This means the block is only at the WAF layer and can be easily fixed.
• Sitemap available with 12 URLs — A sitemap at /sitemaps-1-sitemap.xml lists 12 URLs, helping crawlers discover site pages.
• Sophisticated enterprise SaaS stack — DNS records show integrations with Salesforce, Pendo, Notion, Zoom, Adobe, Atlassian, Docker, Dropbox, and others, indicating a mature tech infrastructure.
• Site links to high-authority press coverage — The site references Bloomberg, Forbes, Reuters, and CoinDesk articles, providing external credibility signals.