putnamauto.com — AI Site Grade

Analysis

Crawler Access: A Single Backdoor for Anthropic

The site is served behind Akamai (AkamaiGHost) and returns 403 Access Denied to every standard browser and nearly every AI crawler tested — GPTBot, Google-Extended, PerplexityBot, ClaudeBot, and ChatGPT-User all receive the same Akamai block. The sole exception is anthropic-ai/1.0, which receives a 200 with 459KB of full HTML served from an nginx origin behind the Akamai wall. This means Anthropic's training crawler has a privileged path to the site's content that no other major AI crawler — including Anthropic's own ClaudeBot — can reach. The robots.txt and llms.txt both return 403, making them invisible to compliant crawlers. No sitemap is accessible. The non-www domain (putnamauto.com) refuses connections entirely.

Cold-Knowledge Gap: The Model Knows a Different Brand

The LLM's cold knowledge describes "Putnam Auto Group" as a luxury dealership group selling BMW, Mercedes-Benz, Audi, and Porsche — with a "no-haggle" pricing model and locations in San Francisco, Oakland, and San Mateo. The actual site reveals a very different business: Putnam Family Dealerships sells 18 mass-market and near-luxury makes (Kia, Ford, Chevrolet, Honda, Toyota, Jeep, Subaru, Mazda, Volkswagen, Volvo, Cadillac, GMC, Chrysler, Dodge, Ram, Fiat, Lexus) from locations in Burlingame, San Carlos, Redwood City, and Oakland — not San Francisco or San Mateo. The model's claimed "over 50 years" aligns with the site's "over 52 years," but the luxury positioning, no-haggle claim, and specific brand mix are fabrications. The model has no awareness of the actual brand portfolio or the "Driven By You" tagline.

Content Architecture: SPA Shell with No Structured Data

The homepage is a single-page application built on the Dealer.com (DDC) platform, loading inventory data via JavaScript bundles. The anthropic-ai bot receives the full rendered page, but the HTML contains zero JSON-LD schema of any type — no AutoDealer, Organization, Product, or Vehicle markup. The page has a single <title> tag and meta description but no Open Graph or Twitter Card tags. Heading structure is flat: one <h1> ("Putnam Automotive") followed by hundreds of <h3> elements listing individual vehicle models. No FAQ, comparison, or structured answer-format signals exist. The site's internal navigation paths (/new-inventory/index.htm, /about-us, etc.) return 404 when accessed directly, confirming the SPA routing dependency.

External Signals: A Digital Ghost

Web searches for "Putnam Family Dealerships," "Putnam Automotive," and individual location queries return zero indexed results from any search engine. No external reviews, Reddit threads, press mentions, or dealer directory listings surfaced. The Wayback Machine shows the site was actively crawled as recently as February 2025 with full content, but the current Akamai configuration has effectively removed the domain from the public web. The site has a Google Search Console verification token in DNS but no visible search presence. This complete absence of external signals means AI models have no secondary sources to correct the hallucinated brand profile they currently hold.

Findings

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

   The site returns 403 Access Denied to GPTBot, Google-Extended, PerplexityBot, ClaudeBot, and ChatGPT-User. Only anthropic-ai/1.0 receives a 200 response with full HTML, creating a privileged backdoor for Anthropic's training crawler.

   What to change: Remove the Akamai block for all major AI crawlers and search engine bots, or implement a consistent allowlist that includes GPTBot, Google-Extended, PerplexityBot, ClaudeBot, and others.

2. robots.txt and sitemap return 403, invisible to compliant crawlers High

   The robots.txt file returns 403 with no user-agent rules, and the sitemap is also blocked (403). This prevents compliant crawlers from discovering allowed paths and content.

   What to change: Serve a publicly accessible robots.txt that allows all major crawlers and link to a valid sitemap.

3. Non-www domain refuses connections entirely Medium

   The non-www domain (putnamauto.com) returns connection refused errors, meaning only the www subdomain is reachable. This can cause confusion and lost traffic.

   What to change: Configure the non-www domain to redirect to the www version or serve the same content.

4. Zero indexed pages in search engines High

   Multiple web searches for the brand name, dealership locations, and site-specific queries returned zero results. The site has no visible search presence despite having a Google Search Console verification token in DNS.

   What to change: Remove the Akamai block for search engine bots and submit a sitemap to Google Search Console and Bing Webmaster Tools.

5. LLM hallucinates luxury brand profile with wrong makes and locations High

   The LLM's cold knowledge describes Putnam Auto Group as a luxury dealership selling BMW, Mercedes-Benz, Audi, Porsche with no-haggle pricing in San Francisco, Oakland, and San Mateo. The actual site sells 18 mass-market makes from Burlingame, San Carlos, Redwood City, and Oakland.

   What to change: Publish accurate structured data (AutoDealer schema) on the site and ensure the site is indexed by search engines to provide authoritative sources for AI models.

6. No JSON-LD schema of any type on the homepage High

   The homepage HTML contains zero JSON-LD schema markup — no AutoDealer, Organization, Product, or Vehicle schema. This deprives AI crawlers and search engines of structured information about the business.

   What to change: Add JSON-LD schema for AutoDealer, Organization, and Vehicle/Product to the homepage and inventory pages.

7. SPA routing causes 404 on direct internal page access High

   Internal navigation paths like /new-vehicles and /about-us return 404 when accessed directly by crawlers, indicating the site relies on JavaScript for routing. This prevents crawlers from discovering content beyond the homepage.

   What to change: Implement server-side rendering or static HTML fallbacks for all internal pages so crawlers can access them without JavaScript.

8. No Open Graph or Twitter Card meta tags Low

   The homepage lacks Open Graph and Twitter Card meta tags, reducing shareability and preview quality on social platforms and AI chat interfaces.

   What to change: Add Open Graph and Twitter Card meta tags with appropriate title, description, and image.

9. Flat heading structure with one h1 and many h3s Low

   The page has a single <h1> and hundreds of <h3> elements, with no <h2> tags. This flat structure reduces semantic clarity for crawlers and accessibility tools.

   What to change: Restructure headings to use a logical hierarchy with h1, h2, and h3 tags.

10. No external reviews, press mentions, or directory listings found Medium

    Web searches for the brand and locations returned zero results from review sites, forums, or news. This absence of external signals means AI models have no secondary sources to correct the hallucinated brand profile.

    What to change: Encourage customer reviews on Google, Yelp, and DealerRater; engage in local press and community events to build online presence.

11. llms.txt returns 403, blocking AI-specific content guidance Medium

    The llms.txt file returns 403, preventing AI crawlers from discovering a curated set of URLs or content hints. This is a missed opportunity to guide AI models to accurate information.

    What to change: Create and serve a publicly accessible llms.txt file that lists key pages and content for AI crawlers.

What's working

• Anthropic training crawler receives full rendered HTML — The anthropic-ai/1.0 bot receives a 200 response with 459KB of full HTML, including inventory data loaded via JavaScript. This ensures Anthropic's training model has access to the site's content.
• Homepage loads inventory data via JavaScript — The SPA homepage successfully loads vehicle inventory data from the Dealer.com platform, providing a rich user experience for human visitors.
• Google Search Console verification token present in DNS — The domain has a Google Search Console verification TXT record, indicating past or intended search engine management.
• Wayback Machine has recent snapshot from February 2025 — The Wayback Machine captured a full snapshot of the site as recently as February 2025, showing the site was previously accessible to crawlers.