patrickcars.com — AI Site Grade

Analysis

The Akamai Firewall That Only Lets Through One AI

Patrickcars.com serves its full homepage, sitemap, and inventory pages to anthropic-ai while returning a 403 Access Denied to every other crawler tested -- including GPTBot, Google-Extended, ClaudeBot, PerplexityBot, and even a standard browser. The site is hosted behind Akamai (edgesuite.net) with a WAF rule that explicitly whitelists the Anthropic crawler user-agent string while blocking everything else, including the Claude-User and Claude-SearchBot entries listed in the robots.txt.

Crawler Access

The robots.txt (accessible only to anthropic-ai) blocks GPTBot, OAI-SearchBot, ChatGPT-User, Claude-User, Claude-SearchBot, PerplexityBot, CCBot, and others from accessing /*.js, /*.css, /*.json, /api/, /apis/, and /pixall/ paths -- but critically does not block these bots from the homepage or content pages. The Akamai WAF overrides this permissiveness entirely, returning 403 before any content is served. The llms.txt exists and is 843 KB -- an unusually large file containing hundreds of inventory page links with descriptions, making it one of the most comprehensive llms.txt files observed. No llms.txt is served to non-Anthropic bots.

Cold-Knowledge Gap

The LLM model describes Patrick Cars as a 20+ dealership network founded in 1980 by Patrick O'Brien, selling Toyota, Honda, Ford, and Chevrolet across the Midwest and Southeast. The actual site tells a different story: the homepage title says "Patrick Dealer Group Schaumburg & Naperville," the HTML class list references mini, hyundai, cadillac, bmw, genesis, land-rover, volvo, jaguar franchises, and the recall page mentions BMW, MINI, Cadillac, Volvo, Jaguar, Land Rover, Hyundai, and Genesis. The site is a multi-franchise luxury and near-luxury dealer group concentrated in the Chicago suburbs, not a broad Midwest/Southeast network. The model's claimed founding story (1980, Patrick O'Brien) does not appear anywhere on the site; the llms.txt says "since 1907."

Schema Posture

The homepage HTML contains no JSON-LD schema in the initial server-rendered markup. The site is built on the Dealer.com (CDK Global) platform (ddc-site CSS class), which loads inventory and structured data entirely through JavaScript after page render. For AI crawlers that cannot execute JS -- which includes all major AI bots except the whitelisted anthropic-ai -- the homepage returns only a 403 shell with 15 words of text. Even anthropic-ai receives a 426 KB HTML document that is predominantly a JavaScript application shell with no visible structured data in the initial payload.

External Signals

Web searches for "Patrick Cars" and related terms returned zero results from any external source -- no reviews, no Reddit threads, no press mentions, no dealer listings. The site has no measurable off-domain footprint visible to search engines. The sitemap contains approximately 3,000+ URLs (455 KB XML file) with lastmod dates of 2026-05-30, suggesting either aggressive future-dating or a platform quirk. The patrick-cars-at-wgn-toy-drive.htm page indicates local community engagement in Chicago, but no external coverage of this or any other activity was found.

Findings

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

   The Akamai WAF returns 403 Access Denied to GPTBot, Google-Extended, ClaudeBot, PerplexityBot, and standard browsers, while allowing only anthropic-ai to access the site. This creates a single-bot visibility monopoly.

   What to change: Update the Akamai WAF rules to allow all major AI crawler user-agent strings (GPTBot, Google-Extended, ClaudeBot, PerplexityBot, etc.) to access the site, or remove the WAF block entirely for known bots.

2. Robots.txt accessible only to anthropic-ai High

   The robots.txt file returns 403 to standard browsers and non-Anthropic bots, but is served to anthropic-ai. This prevents other crawlers from discovering crawl rules.

   What to change: Ensure robots.txt is publicly accessible to all user agents by removing the WAF block for that path.

3. No JSON-LD schema in initial server-rendered HTML High

   The homepage HTML contains no JSON-LD structured data in the initial server-rendered markup. All inventory and structured data is loaded via JavaScript, which AI crawlers that cannot execute JS cannot parse.

   What to change: Add JSON-LD schema (Organization, LocalBusiness, Vehicle, etc.) directly in the server-rendered HTML of the homepage and key pages.

4. Pages render as JavaScript shells for AI crawlers High

   Even for anthropic-ai, the homepage is a 426 KB JavaScript application shell with no visible structured data or content in the initial payload. AI crawlers that cannot execute JS see essentially empty pages.

   What to change: Implement server-side rendering (SSR) or static generation for key pages to deliver meaningful HTML content to crawlers without requiring JavaScript execution.

5. Zero external web presence detected Medium

   Web searches for 'Patrick Cars' and related terms returned zero results from any external source — no reviews, no Reddit threads, no press mentions, no dealer listings. The site has no measurable off-domain footprint.

   What to change: Build external signals through local SEO (Google Business Profile, dealer directories), community event coverage, and press releases to establish off-domain authority.

6. llms.txt not served to non-Anthropic bots Medium

   The llms.txt file (843 KB, one of the largest observed) is only served to anthropic-ai; other AI crawlers receive a 403. This prevents other bots from leveraging the comprehensive inventory listing.

   What to change: Make llms.txt publicly accessible to all user agents by removing the WAF block for that path.

7. Cold knowledge gap: LLM model describes incorrect founding story and dealer network Medium

   The LLM model describes Patrick Cars as a 20+ dealership network founded in 1980 by Patrick O'Brien, selling Toyota, Honda, Ford, Chevrolet across Midwest/Southeast. The actual site shows a luxury multi-franchise group (BMW, MINI, Cadillac, Volvo, etc.) concentrated in Chicago suburbs, with llms.txt stating 'since 1907'.

   What to change: Publish accurate business information (founding year, dealer network size, brands) in structured data and on the homepage to correct the model's knowledge.

8. Sitemap contains future-dated lastmod timestamps Low

   The sitemap XML contains lastmod dates of 2026-05-30, which is in the future. This may confuse crawlers and reduce crawl efficiency.

   What to change: Correct the lastmod dates in the sitemap to reflect actual last modification times.

9. No llms.txt served to non-Anthropic bots Medium

   The llms.txt file is only accessible to anthropic-ai; other AI crawlers receive a 403. This prevents other bots from leveraging the comprehensive inventory listing.

   What to change: Make llms.txt publicly accessible to all user agents by removing the WAF block for that path.

What's working

• Comprehensive llms.txt file with 843 KB of inventory links — The llms.txt file contains hundreds of inventory page links with descriptions, making it one of the most comprehensive observed. This helps AI crawlers discover inventory content.
• Anthropic crawler whitelisted at WAF level — The Akamai WAF explicitly allows the anthropic-ai user agent, ensuring at least one major AI crawler can access the full site content.
• Large sitemap with 3,000+ URLs — The sitemap.xml contains approximately 3,000+ URLs, indicating a comprehensive inventory of pages for crawlers to discover.
• Community engagement page (toy drive) present — The site includes a page about a local toy drive (patrick-cars-at-wgn-toy-drive.htm), demonstrating community involvement that could be leveraged for external signals.