jimellis.com — AI Site Grade

Analysis

The anthropic-ai Backdoor

Jimellis.com operates a two-tier access system where every major AI crawler except anthropic-ai receives a 403 from Akamai, while anthropic-ai bypasses Akamai entirely and hits a separate nginx origin serving full HTML content. This creates a bizarre asymmetry: Claude's crawler gets a 200 with 439KB of homepage content, while GPTBot, Google-Extended, PerplexityBot, and ChatGPT-User all get a 15-word "Access Denied" page.

Crawler Access

The robots.txt (accessible only to anthropic-ai) explicitly names GPTBot, OAI-SearchBot, ChatGPT-User, Claude-User, Claude-SearchBot, and PerplexityBot, but only restricts them from JS, CSS, JSON, and API paths -- it does not block them from the main site. The Akamai WAF enforces the real block at the network layer, returning Akamai's errors.edgesuite.net reference. The non-www domain (jimellis.com) does not resolve at all (connection refused). A 1.9MB llms.txt exists and is served to anthropic-ai, containing hundreds of page listings with descriptions -- an unusually large file that suggests auto-generation from the site's inventory system.

Cold-Knowledge Gap

The LLM model knows Jim Ellis Automotive as a privately held Atlanta group founded in 1971, operating 15+ franchises, and mentions a 2022 class-action lawsuit over deceptive sales practices that settled in 2023. The site itself, via the anthropic-ai-visible pages, presents only awards, executive team bios, and charity work -- no mention of the lawsuit, no crisis communication, no trust-rebuilding content. The model's prior knowledge includes a reputational vulnerability that the site entirely ignores.

Schema Posture

The homepage HTML served to anthropic-ai contains no JSON-LD schema in the head section visible in the snippet. The page uses a DDC (Dealer Digital Commerce) platform from CDK Global, with heavy JavaScript prefetching for inventory facets, filters, and listings. The site is JS-heavy and likely renders most vehicle inventory content client-side, meaning even anthropic-ai may only see the static shell of inventory pages.

External Signals

The model's knowledge of a class-action lawsuit settlement has no corresponding external search results surfaced -- the web search returned zero results for the brand across multiple queries, suggesting either low search-engine visibility or the brand's online footprint is concentrated on individual franchise microsites rather than the corporate domain. The DNS records show CDK Global hosting infrastructure (include:_spf.hosting.cdkglobal.com), confirming the site runs on a standard automotive dealer platform.

Findings

1. Akamai WAF blocks all major AI crawlers except anthropic-ai High

   Every tested AI crawler (GPTBot, Google-Extended, PerplexityBot, ChatGPT-User) receives a 403 Access Denied from Akamai, while anthropic-ai bypasses the WAF and gets a 200 with full HTML content.

   What to change: Remove the Akamai WAF rules that block AI crawlers, or configure the WAF to allow all legitimate AI crawlers.

2. Non-www domain jimellis.com does not resolve Medium

   The non-www version of the domain (jimellis.com) returns a connection refused error, meaning any traffic directed there fails entirely.

   What to change: Configure DNS and web server to redirect jimellis.com to www.jimellis.com.

3. robots.txt is inaccessible to blocked crawlers High

   The robots.txt file returns a 403 for blocked crawlers, preventing them from reading crawl directives. Only anthropic-ai can access it.

   What to change: Ensure robots.txt is publicly accessible and contains correct directives for all crawlers.

4. Sitemap is inaccessible to blocked crawlers High

   The sitemap.xml returns a 403 for blocked crawlers, preventing search engines from discovering the site's pages.

   What to change: Make sitemap.xml publicly accessible.

5. Homepage lacks JSON-LD structured data Medium

   The homepage HTML served to anthropic-ai contains no JSON-LD schema in the head section, missing opportunities for rich search results and AI understanding.

   What to change: Add JSON-LD structured data for organization, local business, and vehicle inventory.

6. Vehicle inventory likely hidden from all crawlers due to client-side rendering High

   The site uses a JS-heavy DDC platform from CDK Global, with inventory content rendered client-side. Even anthropic-ai may only see static shells of inventory pages.

   What to change: Implement server-side rendering or pre-rendering for inventory pages to ensure content is visible to crawlers.

7. No crisis communication or trust-rebuilding content regarding class-action lawsuit Medium

   The site presents awards, executive bios, and charity work but contains no mention of the 2022 class-action lawsuit settlement, leaving a reputational vulnerability unaddressed.

   What to change: Add a page or section addressing the lawsuit, settlement, and steps taken to improve practices.

8. Zero search engine visibility for the corporate domain High

   Multiple web searches for the brand and domain returned zero results, indicating the site is not indexed by major search engines.

   What to change: Resolve crawler access issues and submit sitemap to search engines to begin indexing.

9. llms.txt is unusually large and likely auto-generated Low

   A 1.9MB llms.txt file exists and is served to anthropic-ai, containing hundreds of page listings with descriptions, suggesting auto-generation from the inventory system rather than a curated AI resource.

   What to change: Curate the llms.txt to include only the most important pages and ensure descriptions are accurate and concise.

What's working

• anthropic-ai crawler receives full HTML content — The anthropic-ai crawler bypasses Akamai and receives a 200 response with 439KB of homepage HTML, allowing Claude to index the site's content.
• llms.txt file is served to anthropic-ai — A large llms.txt file (1.9MB) is available and served to anthropic-ai, providing a comprehensive list of pages with descriptions for AI consumption.
• Sitemap is accessible to anthropic-ai — The sitemap.xml is served to anthropic-ai, containing a large number of URLs for discovery.
• robots.txt is accessible to anthropic-ai — The robots.txt file is served to anthropic-ai, allowing it to read crawl directives.
• Awards, executive team, and charity pages are accessible to anthropic-ai — Key brand pages such as awards, executive team bios, and charity work are served to anthropic-ai with full HTML content.