unionpark.com — AI Site Grade

Analysis

Cold-Knowledge Gap: AI models know a Boston financial firm, not this Delaware auto group

A frontier LLM queried cold about "Union Park" returned detailed knowledge of a Boston-based wealth management firm (founded 2005, Tax-Smart Core strategy, Financial Times Top 300 RIA list) — a completely different entity. The actual unionpark.com is a family-owned auto dealership group in Wilmington, Delaware, founded in 1955 by the Ursomarso family, selling BMW, Honda, Volvo, Buick, and GMC across five branded subdomains. The model has zero prior knowledge of the real business.

Crawler Access

The main unionpark.com domain returns 200 with full HTML content to every AI bot tested (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Bytespider, Applebot-Extended, anthropic-ai). No UA-based blocking, no JS shell, no Cloudflare challenge. The site runs on nginx behind Varnish cache with DealerOn as the platform. However, the robots.txt contains zero AI-bot-specific rules — only a catch-all User-Agent: * with a Crawl-delay: 10 and standard disallows for AJAX/print pages. The /llms.txt returns a 404 (DealerOn's generic "Page Not Found" page). The BMW subdomain (unionparkbmw.com) is behind Cloudflare and returns 403 to every bot and browser — completely inaccessible to AI crawlers.

Schema Posture

The main site carries AutomotiveBusiness and WebSite JSON-LD on every page, with department-level AutoDealer, AutoRepair, and AutoBodyShop schemas. The sameAs array is empty on the main domain — no Facebook, Twitter, Instagram, or any social profile linked in the schema. The brand-specific subdomains (Honda, Volvo, Buick GMC) have richer schema including AggregateRating (Honda: 4.0/421 reviews; Volvo: 4.5/240 reviews) and social links. The main site's schema description reads "Union Park Auto Group is a Honda dealer in Wilmington, DE" — understating the five-brand portfolio.

Content & Structure

The homepage is a DealerOn template with 1,119 words of visible text, heavy on keyword-stuffed geographic targeting ("Wilmington, DE and Newark Honda, BMW, GMC, Buick, Volvo customers"). No FAQ schema, no comparison tables, no structured answer-format signals. The "About Us" and "History" pages reveal the family-owned story (founded 1955 as a Pontiac dealership) and the Ursomarso family name. The Community page lists 18 charitable partners and a 2026 GMC Sierra raffle for the Wilmington Firefighters Charity Golf Scramble — a strong local-signal page that AI models would find valuable but have no prior knowledge of.

Fragmented Domain Architecture

The brand operates five separate domains (unionpark.com, unionparkhonda.com, unionparkbmw.com, unionparkbuickgmc.com, unionparkvolvocars.com) with different platforms and access postures. The main site links to the subdomains externally. The BMW subdomain is completely blocked by Cloudflare — no bot or browser can access it. The Honda and Volvo subdomains use a different CMS than the main DealerOn site, with richer schema including aggregate ratings. This fragmentation means an AI crawler following the main site's external links to unionparkbmw.com hits a 403 wall.

Findings

1. BMW subdomain returns 403 to all bots and browsers High

   The unionparkbmw.com subdomain is behind Cloudflare and returns 403 to every AI bot tested, including GPTBot, ClaudeBot, and Google-Extended. Even a regular browser receives a 403. This makes the entire BMW inventory and content invisible to AI crawlers.

   What to change: Remove the Cloudflare block for AI crawlers on unionparkbmw.com, or serve the BMW content from the main unionpark.com domain.

2. No web search results for the dealership or its brands High

   Multiple web searches for 'Union Park Auto Group', 'Union Park Honda', 'unionparkhonda.com', and related terms returned zero results. The dealership has no indexed web presence beyond its own domains, which limits discoverability by AI models relying on web search.

   What to change: Build citations and backlinks from local directories, news articles, and social media to establish a web presence beyond the owned domains.

3. AI models know a Boston financial firm, not the Delaware auto group High

   A frontier LLM queried cold about 'Union Park' returned detailed knowledge of a Boston-based wealth management firm, not the actual Wilmington auto dealership. The model has zero prior knowledge of the real business.

   What to change: Publish authoritative content about the dealership's history, brands, and community involvement to build a distinct AI knowledge profile.

4. Main site schema has empty sameAs array Medium

   The JSON-LD on unionpark.com includes an empty sameAs array, meaning no social media profiles (Facebook, Instagram, Twitter) are linked in the schema. This reduces the site's entity authority and cross-referencing signals for AI models.

   What to change: Add the dealership's social media URLs to the sameAs array in the JSON-LD on all pages.

5. Five separate domains fragment brand presence Medium

   The dealership operates five separate domains (unionpark.com, unionparkhonda.com, unionparkbmw.com, unionparkbuickgmc.com, unionparkvolvocars.com) with different platforms and access postures. AI crawlers following external links from the main site to the BMW subdomain hit a 403 wall.

   What to change: Consolidate all brand content under the main unionpark.com domain, or ensure all subdomains are equally accessible to AI crawlers.

6. No /llms.txt file for AI guidance Medium

   The /llms.txt endpoint returns a 404 (DealerOn's generic 'Page Not Found' page). This means AI models have no structured guidance on which pages to crawl or how to interpret the site's content.

   What to change: Create an /llms.txt file that lists key pages (About, History, Community, inventory pages) and provides a brief description of the dealership.

7. No FAQ or structured answer schema on key pages Medium

   The site lacks FAQ schema, comparison tables, or other structured answer-format signals that AI models use to extract quick answers. This limits the site's ability to appear in AI-generated answers.

   What to change: Add FAQ schema to pages that answer common customer questions (e.g., service hours, financing, inventory).

8. Robots.txt lacks AI-specific directives Low

   The robots.txt contains only a catch-all User-Agent: * with a Crawl-delay: 10 and standard disallows. No AI bot is explicitly allowed or disallowed, leaving crawlers to rely on default behavior.

   What to change: Add explicit rules for AI bots (e.g., GPTBot, ClaudeBot) to ensure they are allowed and to set appropriate crawl delays.

9. Schema description understates brand portfolio Low

   The main site's schema description reads 'Union Park Auto Group is a Honda dealer in Wilmington, DE', which omits the BMW, Volvo, Buick, and GMC brands. This misrepresents the dealership's scope to AI models.

   What to change: Update the schema description to reflect all five brands, e.g., 'Union Park Auto Group is a multi-brand dealership in Wilmington, DE, selling BMW, Honda, Volvo, Buick, and GMC vehicles.'

What's working

• Main site returns full HTML to all AI bots — The unionpark.com domain returns 200 with full HTML content to every AI bot tested, including GPTBot, ClaudeBot, and PerplexityBot. No UA-based blocking or JavaScript shells are present.
• AutomotiveBusiness and department-level schemas in place — Every page on the main site carries AutomotiveBusiness and WebSite JSON-LD, with department-level AutoDealer, AutoRepair, and AutoBodyShop schemas. This provides a solid foundation for entity recognition.
• Honda and Volvo subdomains include AggregateRating schema — The unionparkhonda.com and unionparkvolvocars.com subdomains have richer schema including AggregateRating (Honda: 4.0/421 reviews; Volvo: 4.5/240 reviews) and social links, which enhance credibility signals.
• Community page lists charitable partnerships and events — The Community page details 18 charitable partners and a 2026 GMC Sierra raffle for the Wilmington Firefighters Charity Golf Scramble, providing strong local relevance signals that AI models can use.
• About Us and History pages tell the family-owned story — The About Us and History pages reveal the dealership's founding in 1955 by the Ursomarso family, providing rich narrative content that AI models can use to build entity knowledge.