66north.com — AI Site Grade

Analysis

The site has a sophisticated AI-agent commerce protocol (llms.txt, agents.md, UCP) that is entirely unreachable — every non-homepage URL returns a Cloudflare 403 "Verify your connection" wall to all AI crawlers and browsers alike.

Crawler Access

The homepage at https://66north.com returns 200 with full content to every AI crawler tested (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Bytespider, Applebot-Extended, anthropic-ai) — all receive the same ~1.4MB payload as a browser. However, every other URL on the domain — product pages (/products/dyngja-shell-jacket), the about page (/pages/about), the blog (/blogs/journal), the sitemap (/sitemap_agentic_discovery.xml), the UCP endpoint (/.well-known/ucp), and even the JSON API (/products/dyngja-shell-jacket.json) — returns a Cloudflare 403 "Verify your connection" challenge to all UAs, including browsers. The site is hosted on Shopify behind Cloudflare with a managed challenge rule that blocks deep-page access. The robots.txt is a standard Shopify template with no AI-bot-specific directives — no mention of GPTBot, ClaudeBot, or any AI crawler by name.

llms.txt and Agent Protocol — Present but Hollow

The site has an exemplary llms.txt (and a mirror at /agents.md) that documents the Universal Commerce Protocol (UCP), Shop skill integration, product JSON endpoints, and a full agent shopping flow. This is ahead of nearly all ecommerce sites. But the UCP discovery endpoint (/.well-known/ucp) and the MCP endpoint (/api/ucp/mcp) both return Cloudflare 403 to every bot. The llms.txt tells agents to use GET /products/{handle}.json for product data — but that endpoint also returns 403. The agent protocol is documented but non-functional from any crawler's perspective.

Cold-Knowledge Gap

The LLM knows 66°North as an Icelandic outerwear brand founded in 1926, famous for supplying the Coast Guard and search-and-rescue teams, with iconic products like the "Hells Jacket" and "Kría Parka." It mentions flagship stores in London and Copenhagen and a push into luxury outdoor fashion. The actual site homepage does not mention the Hells Jacket or Kría Parka by name at all — the hero products are the Hornstrandir GORE-TEX Pro Jacket, Dyngja Shell Jacket, Snæfell Shell Jacket, and Tindur Gore-Tex Pro Jacket. The brand's rescue-team heritage is not visible on the homepage; the closest narrative is "Designed and proven in Iceland since 1926" and a "100 years. 100 people." campaign. The cold LLM knowledge is stale by several product cycles and references models the site no longer features.

Schema and Content Posture

The homepage has basic Organization and WebSite schema with SearchAction but no Product schema, no BreadcrumbList, no FAQPage. The site has no FAQ, comparison tables, or structured answer-format content. The blog (/blogs/journal) is entirely blocked behind Cloudflare, meaning AI crawlers cannot access any editorial or brand-story content. The sitemap indexes ~590+ product URLs across English and Icelandic locales, but all are inaccessible to crawlers.

External Signals

The site has zero indexed external mentions discoverable via search — no Reddit threads, no press articles, no review sites surfaced. DNS records show verification tokens for Google, Facebook, Pinterest, Klaviyo, and Apple, indicating a mature marketing stack, but the public web footprint is unusually sparse for a brand of this age and reputation.

Findings

1. Cloudflare 403 challenge blocks all non-homepage URLs from AI crawlers High

   Every URL beyond the homepage — product pages, about page, blog, sitemap, JSON API, UCP endpoint — returns a Cloudflare 403 'Verify your connection' challenge to all AI crawlers and browsers. The homepage is the only accessible page.

   What to change: Remove or relax the Cloudflare managed challenge rule that blocks deep-page access. Allow AI crawler user agents (GPTBot, ClaudeBot, etc.) to bypass the challenge, or serve static HTML versions of key pages to bots.

2. llms.txt and agent protocol documented but non-functional for crawlers High

   The site has an exemplary llms.txt and agents.md that document the Universal Commerce Protocol, Shop skill integration, and product JSON endpoints. However, the UCP discovery endpoint and all referenced JSON endpoints return Cloudflare 403 to every bot, making the protocol unusable.

   What to change: Ensure the UCP endpoint and product JSON endpoints are accessible to AI crawlers by whitelisting their user agents in Cloudflare or serving static responses.

3. LLM cold knowledge is stale by several product cycles Medium

   The LLM knows 66°North as featuring the 'Hells Jacket' and 'Kría Parka', but the homepage hero products are the Hornstrandir GORE-TEX Pro Jacket, Dyngja Shell Jacket, Snæfell Shell Jacket, and Tindur Gore-Tex Pro Jacket. The brand's rescue-team heritage is not prominently featured on the homepage.

   What to change: Update the homepage to feature current hero products and prominently display the brand's rescue-team heritage to align with LLM knowledge and improve AI visibility.

4. Homepage lacks Product schema and structured content Medium

   The homepage has basic Organization and WebSite schema with SearchAction but no Product schema, BreadcrumbList, or FAQPage. The site has no FAQ, comparison tables, or structured answer-format content that AI crawlers can easily parse.

   What to change: Add Product schema for featured products, BreadcrumbList, and FAQPage schema where applicable. Create FAQ or comparison content to provide structured answers.

5. Blog and editorial content entirely blocked from AI crawlers Medium

   The blog at /blogs/journal returns Cloudflare 403 to all crawlers, preventing AI agents from accessing brand-story and editorial content that could inform AI responses.

   What to change: Allow AI crawlers to access the blog by relaxing Cloudflare rules or serving static HTML versions.

6. Zero indexed external mentions found via search Medium

   Web searches for the brand and site returned no results from Reddit, press articles, or review sites. The public web footprint is unusually sparse for a brand of this age and reputation.

   What to change: Investigate why external mentions are not indexed; consider improving SEO and link-building to increase discoverability.

7. robots.txt lacks directives for AI crawlers Low

   The robots.txt is a standard Shopify template with no mention of GPTBot, ClaudeBot, or any AI crawler by name. While not blocking them, it misses the opportunity to guide AI crawlers to accessible content.

   What to change: Add explicit directives for AI crawlers, allowing access to the homepage and any other accessible pages, and disallowing blocked sections.

What's working

• Homepage returns full content to all AI crawlers — The homepage at 66north.com returns a 200 status with full ~1.4MB content to every AI crawler tested, including GPTBot, ClaudeBot, and PerplexityBot.
• Exemplary llms.txt and agent protocol documentation — The site has a well-structured llms.txt and agents.md that document the Universal Commerce Protocol, Shop skill integration, and product JSON endpoints, demonstrating advanced AI-readiness.
• Mature marketing stack with verification tokens — DNS records show verification tokens for Google, Facebook, Pinterest, Klaviyo, and Apple, indicating a sophisticated marketing infrastructure.
• Comprehensive sitemap indexing 590+ product URLs — The sitemap indexes approximately 590 product URLs across English and Icelandic locales, providing a complete inventory of the site's product pages.