dornbracht.com — AI Site Grade

Analysis

Dornbracht's AI crawler access is fully open — every major bot gets a 200 with full content — yet the site has zero structured product schema on its product pages and no llms.txt, creating a paradox where AI engines can read everything but have nothing semantically structured to extract.

Crawler Access

All 11 tested AI bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Bytespider, Applebot-Extended, anthropic-ai, Perplexity-User) receive identical 200 responses with the same 538KB payload as a browser. The robots.txt is a single wildcard rule disallowing only */search and */dashboard — no AI-specific directives exist. The site runs on Azure Front Door (x-azure-ref, x-cache headers) with a Nuxt.js frontend, and the homepage delivers ~491 words of visible text from a plain GET, confirming no JS-rendering barrier. The llms.txt at the root returns a brotli decoding error — the file may exist but is served with a content-encoding the tool cannot decode, effectively making it invisible to AI crawlers that expect plain text.

Schema Posture

The homepage carries Organization and WebSite JSON-LD with correct address, telephone, and social profiles. However, product pages have zero structured data — the COYA single-lever basin mixer page, the TARA design series page, and the COYA design series page all return no Product or ItemPage schema. The FAQ page at /en-gb/service/help-and-faqs lists 20+ questions across 8 categories but uses no FAQPage schema. The press release about Meta and Lulu anniversaries has no NewsArticle or BlogPosting markup. This means AI engines crawling the site get rich prose but no semantic hooks to extract pricing, SKUs, dimensions, or Q&A pairs.

Cold-Knowledge Gap

The LLM prior knows Dornbracht as a German luxury fittings manufacturer with the Tara, Vaia, and Lulu collections, based in Iserlohn. The site itself prominently features COYA (a new 2024/2025 design series) as its hero product — but the cold model did not mention COYA at all. The site also emphasizes sustainability (German Sustainability Award 2026, ReCrafted circular model, FlowReduce water-saving tech) and the "Inspiring your vision" brand claim, none of which surfaced in the model's prior. The gap is that AI engines know the brand's heritage (Tara as a 1990s icon) but are missing the current product narrative — COYA, sustainability credentials, and the Atelier customization program.

External Signals

DNS records show Microsoft 365 mail, Dynamics 365 Marketing (d365mktkey), and Azure infrastructure. The site references a separate ReCrafted shop at dornbracht-recrafted.shop. No Reddit threads or major third-party review aggregators appeared in search results — the brand's external footprint is dominated by its own press releases and reference projects (luxury hotels, yachts, private residences). The sitemap index at /sitemap.xml returns 404 — sitemaps are only accessible via locale-specific paths like /en-gb/sitemap.en-gb.xml (427 URLs), which is functional but means the root-level discovery path fails for crawlers that only check the standard location.

Findings

1. Product pages lack any structured data markup High

   Product pages for COYA and TARA series, as well as individual product detail pages, contain no Product or ItemPage JSON-LD schema. AI crawlers receive rich prose but no semantic hooks for pricing, SKUs, or dimensions.

   What to change: Add Product schema with name, description, SKU, price, and dimensions to all product pages. Use ItemPage schema for design series overview pages.

2. FAQ page has no FAQPage schema High

   The help and FAQs page lists over 20 questions across 8 categories but uses no FAQPage structured data, preventing AI engines from extracting Q&A pairs for rich results.

   What to change: Add FAQPage schema with Question/Answer markup for each FAQ entry.

3. llms.txt returns brotli decoding error High

   The llms.txt file at the root returns a brotli decoding error, likely due to incorrect content-encoding. AI crawlers expecting plain text cannot read it, effectively making the file invisible.

   What to change: Serve llms.txt as plain text without brotli compression, or ensure the Content-Encoding header matches the actual encoding.

4. Press releases lack NewsArticle schema Medium

   The press release about the Meta and Lulu anniversary has no NewsArticle or BlogPosting structured data, reducing its visibility in AI-powered news and search features.

   What to change: Add NewsArticle schema with headline, datePublished, author, and articleBody to press release pages.

5. Root sitemap.xml returns 404 Medium

   The standard sitemap index at /sitemap.xml returns a 404 error. Sitemaps are only available via locale-specific paths, which may cause crawlers that only check the standard location to miss the sitemap.

   What to change: Create a root sitemap index at /sitemap.xml that points to all locale-specific sitemaps.

6. AI models lack awareness of current product narrative Medium

   The LLM prior did not mention COYA, the hero product series, nor sustainability credentials like the German Sustainability Award or ReCrafted program. AI engines know the brand's heritage but miss the current product story.

   What to change: Ensure product pages and sustainability content are well-structured with schema and prominently linked from the homepage to improve AI knowledge extraction.

7. Robots.txt has no AI-specific directives Low

   The robots.txt only disallows */search and */dashboard with a wildcard rule. No AI-specific directives exist, which is fine for access but means the site cannot signal preferred crawl behavior for AI bots.

   What to change: Consider adding explicit directives for AI bots (e.g., GPTBot, ClaudeBot) to allow or disallow specific paths if needed.

8. Limited external review presence on third-party sites Low

   No Reddit threads or major third-party review aggregators appeared in search results. The brand's external footprint is dominated by its own press releases and reference projects, which may limit AI signals from user-generated content.

What's working

• All major AI bots receive full content access — All 11 tested AI bots receive identical 200 responses with full HTML content. No bot-specific blocking or JS-rendering barriers exist.
• Homepage has correct Organization and WebSite schema — The homepage includes Organization and WebSite JSON-LD with accurate address, telephone, and social profile URLs, providing a solid foundation for brand knowledge.
• Fast server response with Azure Front Door CDN — The site uses Azure Front Door for CDN and caching, delivering fast responses. Homepage loads in under 1 second with full content.
• Locale-specific sitemaps are functional and well-structured — Sitemaps like /en-gb/sitemap.en-gb.xml are accessible and contain 427 URLs, ensuring crawlers can discover content via locale paths.
• Product and content pages have substantial text content — Product pages and press releases contain 300-1000 words of descriptive text, providing rich material for AI crawlers to index.
• AI models have strong prior knowledge of the brand — The LLM prior correctly identifies Dornbracht as a German luxury fittings manufacturer with key collections (Tara, Vaia, Lulu) and location in Iserlohn.