andersonautomotivegroup.com — AI Site Grade

Analysis

---

The site is fully live and served to anthropic-ai while returning 403 Access Denied to every other AI crawler and all browser traffic

The Akamai WAF at www.andersonautomotivegroup.com blocks every user-agent tested — GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, Claude-SearchBot, PerplexityBot, Perplexity-User, Google-Extended, Bytespider, Applebot-Extended, and even a standard browser — all returning a 403 Akamai "Access Denied" page. The sole exception is anthropic-ai (the training crawler), which receives a full 200 response with 355KB of HTML served from an nginx origin behind the Akamai edge. This means Claude's training pipeline sees the entire site, but no other AI engine — including Claude's own search/reasoning bots (ClaudeBot, Claude-User, Claude-SearchBot) — can access a single page.

Crawler Access

The robots.txt (only readable by anthropic-ai) explicitly disallows GPTBot, OAI-SearchBot, ChatGPT-User, Claude-User, Claude-SearchBot, PerplexityBot, CCBot, and others from /api/, /apis/, /pixall/, */pixall/, and all .js, .css, .json files — but does not block anthropic-ai at all. The llms.txt file exists and is served (1.6MB) but is also only accessible to anthropic-ai. The sitemap.xml contains over 1.1MB of URLs with lastmod dates as recent as 2026-05-30, indicating the site is actively maintained behind the WAF wall. No Google-Extended or Applebot-Extended rules exist in robots.txt.

Cold-Knowledge Gap

A frontier LLM queried cold about "Anderson Automotive Group" described it as a privately held group based in North Carolina selling BMW, Mercedes-Benz, Honda, and Toyota, founded by Billy Anderson in the 1970s with over 10 dealerships. The actual site (visible via Wayback and the anthropic-ai fetch) reveals a much larger operation: 20+ dealerships across North Carolina and South Carolina under the "Fred Anderson" brand, selling Toyota, Nissan, Kia, Honda, Acura, Hyundai, Genesis, Subaru, Chevrolet, Buick, GMC, Cadillac, Mazda, and Infiniti — but no BMW or Mercedes-Benz presence. The cold model hallucinated luxury German brands and the founder's name, while missing the actual "Fred Anderson" naming convention and the full geographic footprint (Asheville, Raleigh, Sanford, Greenville, Greer, Charleston).

Schema Posture

The homepage and subpages contain zero JSON-LD structured data of any kind. No AutoDealer, Organization, LocalBusiness, Product, FAQPage, or BreadcrumbList schema is present. The site runs on the Dealer.com (DDC) platform, which typically injects schema, but the Akamai WAF is stripping or blocking it for non-anthropic-ai traffic. The visible HTML includes rich dealer inventory data (franchise classes, location coordinates, phone numbers) that would be highly valuable for AI answer engines if exposed via schema.

External Signals

The brand has near-zero external web footprint discoverable via search. No Reddit threads, no review aggregator pages, no press mentions, and no independent articles surfaced in multiple search queries. The cold LLM's mention of "mixed reviews on Google and DealerRater" could not be verified through any indexed source. The site's own llms.txt lists pages for "News & Events" and "Drive for a Difference" that return empty descriptions, suggesting thin or unpopulated content.

Findings

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

   The Akamai WAF at www.andersonautomotivegroup.com returns 403 Access Denied to every tested user-agent (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, Claude-SearchBot, PerplexityBot, Perplexity-User, Google-Extended, Bytespider, Applebot-Extended, and standard browser) except anthropic-ai, which receives a full 200 response. This means Claude's training pipeline sees the entire site, but no other AI engine can access any page.

   What to change: Reconfigure the Akamai WAF to allow access for all major AI crawlers (GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, etc.) or serve a static HTML version to bots.

2. robots.txt disallows multiple AI crawlers from key paths High

   The robots.txt file, only readable by anthropic-ai, explicitly disallows GPTBot, OAI-SearchBot, ChatGPT-User, Claude-User, Claude-SearchBot, PerplexityBot, CCBot, and others from /api/, /apis/, /pixall/, */pixall/, and all .js, .css, .json files. No rules exist for Google-Extended or Applebot-Extended.

   What to change: Remove disallow rules for AI crawlers from robots.txt, or ensure the WAF allows access to pages blocked by robots.txt for compliant crawlers.

3. llms.txt file only accessible to anthropic-ai Medium

   The llms.txt file (1.6MB) exists and is served, but only anthropic-ai can access it; all other user-agents receive a 403. This prevents other AI engines from using the file for context.

   What to change: Allow public access to llms.txt by removing the WAF block for that path.

4. Zero JSON-LD structured data on homepage and subpages High

   The homepage and subpages contain no JSON-LD structured data of any kind. No AutoDealer, Organization, LocalBusiness, Product, FAQPage, or BreadcrumbList schema is present. The visible HTML includes rich dealer inventory data that would be valuable for AI answer engines if exposed via schema.

   What to change: Add JSON-LD structured data for AutoDealer, Organization, LocalBusiness, and Product schema on all relevant pages.

5. Cold LLM hallucinates luxury brands and founder name High

   A frontier LLM queried cold about 'Anderson Automotive Group' hallucinated BMW and Mercedes-Benz dealerships and the founder as 'Billy Anderson', while missing the actual 'Fred Anderson' naming convention and the full geographic footprint (20+ dealerships across NC and SC).

   What to change: Publish authoritative content (About, History pages) with structured data and ensure AI crawlers can access it to correct the knowledge gap.

6. Near-zero external web footprint discoverable via search Medium

   Multiple web searches for the brand, reviews, and dealership mentions returned zero indexed results. No Reddit threads, review aggregator pages, press mentions, or independent articles were found.

   What to change: Build external signals through press releases, local business listings, review profiles, and social media engagement.

7. News & Events and Drive for a Difference pages have empty descriptions in llms.txt Low

   The llms.txt file lists pages for 'News & Events' and 'Drive for a Difference' that return empty descriptions, suggesting thin or unpopulated content.

   What to change: Populate these pages with substantive content or remove them from llms.txt.

8. Sitemap.xml accessible only to anthropic-ai High

   The sitemap.xml (1.1MB) is served to anthropic-ai but returns 403 to all other user-agents, preventing other search engines and AI crawlers from discovering the site's URL structure.

   What to change: Allow public access to sitemap.xml by removing the WAF block for that path.

9. Browser traffic blocked by WAF High

   Standard browser user-agents receive a 403 Access Denied from the Akamai WAF, meaning human visitors cannot access the site without special configuration.

   What to change: Reconfigure the WAF to allow standard browser traffic, or ensure the site is accessible to the public.

What's working

• Site fully served to anthropic-ai training crawler — The anthropic-ai crawler receives a full 200 response with 355KB of HTML, ensuring Claude's training pipeline has complete access to the site content.
• llms.txt file exists and is populated (1.6MB) — The site has a large llms.txt file (1.6MB) that lists many pages, indicating an effort to provide AI-friendly content, even though it is currently blocked to most crawlers.
• Sitemap.xml is large (1.1MB) and current — The sitemap.xml contains over 1.1MB of URLs with lastmod dates as recent as 2026-05-30, indicating the site is actively maintained and has a comprehensive URL inventory.
• robots.txt exists and is customized — The robots.txt file is present and contains specific rules for various crawlers, showing intentional management of crawler access.
• Wayback Machine snapshot shows historical content — A Wayback Machine snapshot from January 2025 shows the site was publicly accessible with rich content about the dealership group.