pieinsurance.com — AI Site Grade

Analysis

I have enough data to write a thorough audit. Let me compile my findings.

Pie Insurance — AI-Visibility Audit

The site has an anthropic-domain-verification TXT record in DNS proving explicit Claude crawler clearance, yet zero structured data (JSON-LD schema) on any page tested — a striking contradiction for a data-driven insurtech that cold LLM knowledge already describes accurately without needing live retrieval.

Crawler Access

All eleven AI bot user-agents tested (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, anthropic-ai, Bytespider, Applebot-Extended, Perplexity-User) return HTTP 200 with full content from the homepage — identical byte sizes (~152KB) to a browser baseline. The site runs on Vercel (Next.js) with no WAF blocking, no Cloudflare challenge, and no UA-based gating. The robots.txt is a single Allow: / for User-Agent: * with zero AI-specific directives — no bot is restricted, but none is explicitly welcomed either. The llms.txt returns HTTP 404 (a Next.js error page), meaning no AI-friendly content map exists despite the domain being anthropic-verified.

Cold-Knowledge Gap

The LLM cold-knowledge snapshot of Pie Insurance is surprisingly accurate and current: it correctly names founders John Swigart and Dax Craig, the 2017 founding, the $300M+ funding total including the $118M Series D in 2022, the A- AM Best rating, and the Trustpilot reputation. This is unusual — most insurtech brands have stale or hallucinated cold knowledge. The gap is not in what the model knows but in what the site fails to reinforce: the site never mentions the founders, the funding history, the Series D, or the AM Best rating on any fetched page. The homepage and product pages are purely functional (quote flow, coverage descriptions) with no brand-story or trust-signal content that would anchor the model's positive prior.

Schema Posture

Zero JSON-LD schema was found on any page tested — homepage, workers-comp, BOP, FAQ, blog, state pages, contact page. No Organization, Product, FAQPage, BreadcrumbList, LocalBusiness, or WebSite schema exists. This is a critical miss for an insurance carrier that wants AI engines to confidently extract structured facts (coverage areas, states served, NAIC number, AM Best rating). The FAQ page has visible Q&A content but no FAQPage markup. The blog has 332 posts with no Article or BlogPosting schema.

Content & Structure

The site has strong editorial content — the workers-comp FAQ page runs 760 words of genuine Q&A, the Texas state page runs 1,348 words of state-specific guidance, and the blog has 332 posts with substantive articles (the AI workplace safety post runs 2,066 words). The homepage is thin at 468 words but functional. The sitemap contains 852 URLs, many of which are thin state-specific workplace-notice pages (40+ state pages with boilerplate content). The site uses Next.js with client-side rendering but all tested pages returned full HTML to crawlers — no JS-shell risk detected.

External Signals

The DNS TXT records reveal a sophisticated vendor stack: Anthropic domain verification (explicit Claude clearance), multiple Google site verifications, Stripe, HubSpot, Salesforce, Greenhouse (ATS), Intacct, Mixpanel, MongoDB, and Cursor verifications. This is a well-resourced insurtech with a mature tech stack. No Reddit threads or negative review signals surfaced in search. The copyright footer reads "2026" — a forward-dated year that may confuse temporal reasoning in AI models.

Findings

1. Zero JSON-LD schema on any page High

   No structured data (Organization, Product, FAQPage, BreadcrumbList, etc.) was found on any tested page, including the homepage, product pages, FAQ, blog, and contact page. This prevents AI engines from confidently extracting structured facts like coverage areas, NAIC number, and AM Best rating.

   What to change: Add JSON-LD schema for Organization, WebSite, BreadcrumbList, Product, FAQPage, and Article/BlogPosting across all relevant pages.

2. Missing llms.txt file Medium

   The llms.txt file returns a 404 error, meaning no AI-friendly content map exists despite the domain being anthropic-verified. This is a missed opportunity to guide AI crawlers to key resources.

   What to change: Create an llms.txt file listing key pages (homepage, product pages, FAQ, blog, state pages) to provide a structured entry point for AI crawlers.

3. Missing brand story and trust signals on key pages Medium

   The site never mentions founders, funding history, Series D, or AM Best rating on any fetched page. The homepage and product pages are purely functional, lacking the brand-story content that would reinforce the model's positive cold knowledge.

   What to change: Add a dedicated 'About' or 'Why Pie' section on the homepage and product pages that includes founding story, funding milestones, and AM Best rating.

4. Copyright footer forward-dated to 2026 Low

   The copyright footer reads '2026', which may confuse temporal reasoning in AI models and reduce trust in the site's timeliness.

   What to change: Update the copyright year to the current year or use a dynamic year.

5. Thin state-specific workplace-notice pages Low

   The sitemap contains 852 URLs, many of which are thin state-specific workplace-notice pages with boilerplate content. These may dilute crawl budget and provide little value to AI crawlers.

   What to change: Consolidate or enrich thin state pages with unique, substantive content to improve crawl efficiency and AI value.

What's working

• Explicit Anthropic domain verification — The DNS TXT records include an anthropic-domain-verification record, giving Claude explicit clearance to crawl the site.
• All 11 AI bots receive full content — All tested AI bot user-agents return HTTP 200 with full HTML content, identical to a browser baseline. No WAF, Cloudflare, or UA-based gating is present.
• Accurate cold LLM knowledge of Pie Insurance — The LLM cold-knowledge snapshot correctly names founders, funding history, AM Best rating, and Trustpilot reputation — unusual for an insurtech brand.
• Strong editorial content on FAQ and blog — The workers-comp FAQ page has 760 words of genuine Q&A, the Texas state page has 1,348 words of state-specific guidance, and the blog has 332 substantive posts (e.g., 2,066-word AI workplace safety article).
• Mature tech stack with multiple verifications — DNS TXT records reveal a sophisticated vendor stack including Stripe, HubSpot, Salesforce, Greenhouse, Intacct, Mixpanel, MongoDB, and Cursor verifications, indicating a well-resourced engineering team.
• No negative external signals found — Web searches for reviews, Reddit threads, and AM Best rating did not surface negative or contradictory information.