cloverhealth.com — AI Site Grade

Analysis

I have a comprehensive picture now. Let me compile the audit.

Clover Health — AI-Visibility Audit

The site's robots.txt and llms.txt both return 200 with a full HTML page (not plain text), meaning every AI crawler that requests either file gets served the site's JavaScript shell — a silent crawl-block that no Disallow rule could match.

Crawler Access

All major AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Applebot-Extended, Bytespider — receive 200 with identical byte payload (219 KB) from Google Frontend infrastructure. No UA-based blocking exists. However, the homepage is a heavy JS-rendered single-page application served via Google Cloud Run. The plain GET returns ~2,095 words of visible text, but much of the page's interactive content (plan search, ZIP lookup, provider network) is loaded dynamically and may be invisible to crawlers that do not execute JavaScript. The anthropic-ai bot returned an error (connection issue), not a block.

Infrastructure & Discovery Gaps

The site runs on Google Cloud Run behind Google Frontend with Google DNS and Google-hosted nameservers. sitemap.xml returns 404. robots.txt returns 404 (serving the HTML app shell instead of a text file). llms.txt returns 404 (same HTML shell). No crawl directives exist anywhere. The /blog path enters a redirect loop. Key pages that the navigation links to — /about, /about/press, /leadership — all return 404. The working /about-us page contains zero body content beyond the navigation shell. The /investors page times out. This means AI crawlers cannot discover the site's content structure through standard discovery protocols.

Cold-Knowledge Gap

The LLM prior on Clover Health is dominated by negative reputational signals: the 2021 SPAC merger, Hindenburg Research's short-seller allegations, a Department of Justice investigation, and ongoing financial losses. The site itself contains zero mention of any of these events. The homepage and all pages present Clover purely as a Medicare Advantage provider with $0/month plans, dental/vision/hearing benefits, and the Clover Assistant AI tool. The site's own narrative — "Medicare done differently" — is entirely absent from the cold model's knowledge. The model knows Clover as a controversial, financially struggling tech-insurance hybrid; the site presents Clover as a straightforward Medicare plan with good benefits. This is a dangerous gap: AI engines answering "What is Clover Health?" will surface the Hindenburg/DOJ narrative before the site's own messaging.

Schema & Content Posture

The homepage carries a single WebPage schema with Organization markup (name, address, phone, social profiles). The plans page uses HealthInsurancePlan schema. The FAQ page (/understanding-medicare/medicare-faq) has no schema at all — no FAQPage markup despite containing 12 clearly structured Q&A pairs. The blog article uses Article schema with a datePublished of 2026-03-09 (a future date), which will confuse freshness signals. No Product, MedicalBusiness, or HealthPlanCostEstimate schema types are present. The site has FAQ content and comparison language ("PPO & HMO") but no structured markup to surface these in AI answer boxes.

External Signals

The DNS TXT records confirm OpenAI domain verification (openai-domain-verification=dv-o8wIlbgcFh38vzCIoBo92FxJ) and Anthropic domain verification (anthropic-domain-verification-6m0yhp=...), meaning both companies have verified Clover Health's domain for API/crawler use. The site links to a Medium publication ("Off the Charts") and an investor relations subdomain (investors.cloverhealth.com), but neither is discoverable through the main site's sitemap or robots.txt. No external review sites, Reddit threads, or press coverage were surfaced in search — the site's external citation footprint is essentially zero for AI retrieval purposes.

Findings

1. robots.txt returns HTML app shell instead of plain text High

   The robots.txt file at cloverhealth.com returns a 200 status with a full HTML page (the JavaScript app shell), not a plain text file. Every AI crawler that requests it receives the site's JS shell, effectively blocking crawl discovery without any Disallow rule.

   What to change: Serve robots.txt as a plain text file with appropriate crawl directives for AI bots. Ensure the server returns Content-Type: text/plain for the /robots.txt path.

2. sitemap.xml returns 404 High

   The sitemap.xml file returns a 404 error, meaning AI crawlers cannot discover the site's URL structure through standard sitemap protocols.

   What to change: Generate and serve a valid sitemap.xml listing all important pages, and submit it to Google Search Console and Bing Webmaster Tools.

3. llms.txt returns 404 Medium

   The llms.txt file returns a 404 error, missing an opportunity to provide AI crawlers with a curated list of important pages and context.

   What to change: Create an llms.txt file that lists key pages (e.g., about, plans, FAQ) and provides a brief description of the site for AI crawlers.

4. Homepage is a heavy JS-rendered SPA High

   The homepage is a JavaScript-rendered single-page application served via Google Cloud Run. While the plain GET returns ~2,095 words of visible text, much of the interactive content (plan search, ZIP lookup, provider network) is loaded dynamically and may be invisible to crawlers that do not execute JavaScript.

   What to change: Implement server-side rendering or pre-rendering for key content pages to ensure AI crawlers can access all text content without JavaScript execution.

5. /blog path enters a redirect loop High

   The /blog URL returns an error due to exceeding maximum redirects, making the blog completely inaccessible to crawlers.

   What to change: Fix the redirect loop on the /blog path so that crawlers can access blog content.

6. Key navigation pages return 404 High

   Pages linked from the navigation — /about, /about/press, /leadership — all return 404 errors. The working /about-us page contains zero body content beyond the navigation shell. The /investors page times out.

   What to change: Ensure all navigation-linked pages return 200 with meaningful content. Fix the /about-us page to include body content. Resolve the timeout on /investors.

7. FAQ page lacks FAQPage schema markup Medium

   The Medicare FAQ page contains 12 clearly structured Q&A pairs but has no FAQPage schema markup, missing an opportunity to appear in AI answer boxes and rich snippets.

   What to change: Add FAQPage schema markup to the FAQ page, wrapping each Q&A pair in the appropriate structured data.

8. Blog article has a future datePublished Medium

   A blog article uses a datePublished of 2026-03-09, a future date, which will confuse AI crawlers and search engines regarding content freshness.

   What to change: Correct the datePublished field to the actual publication date. Ensure all dates are accurate and in ISO 8601 format.

9. Cold LLM knowledge is dominated by negative reputational signals High

   The LLM prior on Clover Health is dominated by the 2021 SPAC merger, Hindenburg Research's short-seller allegations, a DOJ investigation, and ongoing financial losses. The site's own narrative — 'Medicare done differently' — is entirely absent from the cold model's knowledge. AI engines answering 'What is Clover Health?' will surface the negative narrative before the site's own messaging.

   What to change: Publish authoritative content (e.g., an 'About' page with the company story, press releases, and financial updates) and ensure it is indexed and crawlable. Consider submitting to Google's Knowledge Graph and using schema markup to reinforce the desired narrative.

10. External citation footprint is near zero Medium

    No external review sites, Reddit threads, or press coverage were surfaced in search for AI retrieval. The site's external citation footprint is essentially zero, limiting its visibility in AI-generated answers that rely on third-party sources.

    What to change: Encourage customer reviews on platforms like Google, Yelp, and Healthgrades. Engage in PR and content marketing to generate positive press coverage and backlinks.

What's working

• OpenAI and Anthropic domain verification in place — DNS TXT records confirm OpenAI and Anthropic domain verification, meaning both companies have verified Clover Health's domain for API/crawler use, enabling potential access for their crawlers.
• Plans page uses HealthInsurancePlan schema — The plans page includes HealthInsurancePlan schema markup, which helps AI crawlers understand the insurance offerings and may enable rich results.
• Homepage has WebPage and Organization schema — The homepage carries WebPage schema with Organization markup including name, address, phone, and social profiles, providing basic structured data for AI crawlers.
• Blog article uses Article schema — A blog article includes Article schema markup, which helps AI crawlers identify and surface blog content.
• No user-agent based blocking of AI crawlers — All major AI crawlers receive 200 responses from the server; there is no explicit UA-based blocking, meaning the site does not intentionally exclude AI bots.