svp.com — AI Site Grade

Analysis

svp.com: A Site That Does Not Exist for AI Crawlers

The domain svp.com hosts Groupe SVP, a French B2B compliance and HR services firm founded in 1935 — not the sewing-machine parent company (SVP Worldwide) that the LLM cold-knowledge model hallucinated. Every single AI crawler tested — GPTBot, ClaudeBot, Google-Extended, PerplexityBot, ChatGPT-User, OAI-SearchBot, Applebot-Extended, Bytespider, and anthropic-ai — receives a 403 from Cloudflare's JavaScript challenge on the homepage, with zero visible text returned.

Crawler Access

The live site is entirely behind Cloudflare's managed challenge (cf-ray headers present, CSP blocking all script sources except Cloudflare's own). No robots.txt is served to crawlers — the request returns the same 403 JS challenge page. The historical robots.txt from July 2024 (retrieved via Wayback Machine) explicitly disallowed GPTBot, ChatGPT-User, and Google-Extended from the entire site via Disallow: /, while allowing all other user-agents. The llms.txt endpoint also returns a 403. The site uses Cloudflare nameservers (paloma.ns.cloudflare.com, stanley.ns.cloudflare.com) and serves from Cloudflare IPs (104.20.20.203, 172.66.172.44).

Content and Schema

The Wayback Machine snapshot from December 2024 reveals a French-language B2B portal offering legal/regulatory information, HR/payroll software (Agrume), economic intelligence consulting, and public-sector services. The homepage H1 reads "SVP, le partenaire de confiance des professionnels des secteurs privés et publics." The site contains zero JSON-LD schema of any type. No FAQ schema, no Organization schema, no BreadcrumbList. Heading structure is flat (H1 → H2 → H3) with no semantic depth. The site has a large content library (news, publications, webinars) but no structured answer-format signals like FAQ blocks or comparison tables.

Cold-Knowledge Gap

The LLM model described svp.com as "the online platform for SVP Worldwide, the parent company of Singer, Viking, and Pfaff sewing machines" — a complete category error. The actual site is Groupe SVP, a French business information and compliance services provider. The model also hallucinated specific product lines (Singer Quantum Stylist, Pfaff Creative Icon) and customer-review signals (Trustpilot, BBB complaints) that have no connection to this domain. The model's secondary knowledge about Groupe SVP France (founded 1935, 100,000+ clients, subscription-based legal/HR helpline) is accurate but entirely absent from the model's primary response about the domain.

External Signals

The domain's MX records point to ppe-hosted.com (a French email security provider). TXT records include google-site-verification tokens, sendinblue-code for email marketing, and Salesforce SPF includes — confirming a French B2B SaaS stack. No external reviews, Reddit threads, or press mentions for svp.com were surfaced by search. The domain has been archived by the Wayback Machine since 1997 with over 2,000 captures, indicating a long-established but low-public-profile operation.

Findings

1. All major AI crawlers receive 403 from Cloudflare JS challenge High

   Every tested AI crawler (GPTBot, ClaudeBot, Google-Extended, PerplexityBot, ChatGPT-User, OAI-SearchBot, Applebot-Extended, Bytespider, anthropic-ai) receives a 403 response from Cloudflare's JavaScript challenge on the homepage, with zero visible text returned.

   What to change: Remove the Cloudflare JS challenge for AI crawler user-agents, or serve a static HTML version of the homepage to bots.

2. Historical robots.txt disallows GPTBot, ChatGPT-User, and Google-Extended High

   The July 2024 robots.txt snapshot explicitly disallows GPTBot, ChatGPT-User, and Google-Extended from the entire site via Disallow: /, while allowing all other user-agents. The current robots.txt endpoint returns a 403, so crawlers cannot even read it.

   What to change: Update robots.txt to allow AI crawlers, or remove the explicit disallow directives for GPTBot, ChatGPT-User, and Google-Extended.

3. llms.txt endpoint returns 403 Medium

   The llms.txt file, which helps AI crawlers discover content, returns a 403 error and is not accessible.

   What to change: Create and serve an llms.txt file listing key content pages for AI crawlers.

4. Zero JSON-LD structured data on the site High

   The site contains no JSON-LD schema of any type — no Organization, FAQ, BreadcrumbList, or other structured data that helps AI models understand and cite content.

   What to change: Add JSON-LD structured data for Organization, BreadcrumbList, and FAQ (if applicable) to all pages.

5. LLM cold knowledge hallucinates SVP Worldwide sewing machines High

   The LLM model incorrectly identifies svp.com as the platform for SVP Worldwide (Singer, Viking, Pfaff sewing machines), when the actual site is Groupe SVP, a French B2B compliance and HR services firm. This hallucination could mislead AI-generated summaries.

   What to change: Publish clear, authoritative content (e.g., About page, Wikipedia entry) that explicitly states the company's identity and services to correct LLM knowledge.

6. No indexed URLs or search results for svp.com High

   Web searches for svp.com and related queries return zero results, and the URL discovery tool found no known URLs. The site is effectively invisible to search engines and AI crawlers.

   What to change: Ensure the site is indexable by search engines by removing the Cloudflare challenge for Googlebot and Bingbot, and submit a sitemap to Google Search Console.

7. Flat heading structure with no semantic depth Low

   The homepage uses a flat H1 → H2 → H3 heading hierarchy with no semantic depth, which reduces content clarity for AI parsers.

   What to change: Use a deeper, more descriptive heading structure (H1, H2, H3, H4) to improve content hierarchy.

8. No FAQ blocks or comparison tables for structured answers Medium

   The site has a large content library but lacks FAQ blocks or comparison tables that help AI models extract direct answers.

   What to change: Add FAQ schema and structured comparison tables to relevant content pages.

What's working

• Domain has been archived since 1997 with over 2,000 captures — The domain has a long history on the web, with Wayback Machine captures dating back to 1997, indicating stability and longevity.
• Large French-language content library with news and publications — The site hosts a substantial library of news, publications, and webinars relevant to French B2B compliance and HR audiences.
• Cloudflare protection with JS challenge mitigates bot abuse — The site uses Cloudflare's managed challenge to block malicious traffic, which protects against DDoS and scraping attacks.
• Email and marketing infrastructure configured (Sendinblue, Salesforce) — TXT records show Sendinblue email marketing and Salesforce SPF includes, indicating a functional B2B SaaS stack.