geha.com — AI Site Grade

Analysis

GEHA.com: Akamai silently blocks nearly all AI crawlers, and zero pages carry structured data

The most consequential finding is that every major AI crawler except anthropic-ai times out against GEHA's Akamai edge, while the robots.txt explicitly allows all bots — creating a silent crawl blockade that neither the site nor the AI engines can diagnose from the robots file alone.

Crawler Access

The robots.txt at geha.com/robots.txt is permissive: User-agent: * with Allow: / and a single sitemap reference. No AI-bot-specific rules exist. However, compare_bot_access testing on the homepage reveals that GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Bytespider, and Applebot-Extended all receive zero-byte responses (read timeouts). Only anthropic-ai (the API crawler, distinct from ClaudeBot) returns a full 200 with the same content as a browser. The site runs on Akamai CDN (confirmed via DNS akam.net nameservers and x-akamai-transformed response header). Akamai's bot management is silently dropping AI crawlers at the edge, not returning a 403 or challenge page — meaning the crawlers see a connection timeout, not a block signal they could interpret.

Content & Schema

Every page examined — homepage, plan overview, plan detail, FAQs, blog, about page, resource center — contains zero JSON-LD structured data of any type. No HealthPlan, Organization, FAQPage, WebSite, or BreadcrumbList schema is present. The homepage has a clear H1 ("Welcome new and returning members"), multiple H2s for plan categories, and rich FAQ content on the /why-geha/customer-care/faqs page (11,533 words of Q&A), but none of it is marked up as FAQPage. The plan comparison page (/plan-compare) uses comparison language and interactive tables but has no Product or HealthPlan schema. The blog (/geha-blog) lists 50+ articles with dates and categories but no BlogPosting or Article schema.

Cold-Knowledge Gap

The LLM's cold knowledge about GEHA is surprisingly accurate and detailed: it knows GEHA serves 2+ million federal employees, was founded in 1937 by postal workers, offers Elevate and Standard Option plans, and mentions J.D. Power satisfaction ratings and low premium increases. The site itself confirms the founding story and plan names but never mentions J.D. Power, member satisfaction survey results, or specific competitive differentiators like "lowest premium increases" — the very facts the model already knows from external sources. The site's own About page is only 439 words and lacks the depth of the model's prior knowledge.

External Signals

The llms.txt at geha.com/llms.txt returns a 404. No AI-friendly content map exists. DNS records show Adobe IDP verification, OneTrust (consent management), and Smartsheet — indicating a complex martech stack but no AI-readiness tooling. The sitemap contains 403 URLs including a suspicious entry (/abo-sfsc-delete-me) that suggests a staging/test page leaked into production. The member portal lives on a separate subdomain (member.mygeha.com) behind a login wall, meaning AI crawlers cannot access the actual claims, ID card, or member resource content that members rely on.

Findings

1. Akamai silently blocks all major AI crawlers except anthropic-ai High

   The site's Akamai CDN drops connections from GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Bytespider, and Applebot-Extended with read timeouts, while robots.txt allows all bots. This creates a silent blockade that AI engines cannot diagnose.

   What to change: Configure Akamai bot management to allow AI crawlers (GPTBot, ClaudeBot, etc.) or serve them static HTML content instead of dropping connections.

2. Zero JSON-LD structured data on any page High

   Every page examined—homepage, plan overview, plan detail, FAQs, blog, about page, resource center—contains no JSON-LD structured data of any type (HealthPlan, Organization, FAQPage, WebSite, BreadcrumbList, etc.).

   What to change: Add JSON-LD structured data: Organization on homepage, HealthPlan on plan pages, FAQPage on the FAQs page, BlogPosting on blog articles, and BreadcrumbList site-wide.

3. llms.txt returns 404 Medium

   The site does not provide an llms.txt file at /llms.txt, missing an opportunity to guide AI crawlers to key content.

   What to change: Create an llms.txt file listing key pages (plan overview, FAQs, blog) to help AI crawlers discover important content.

4. Rich FAQ content lacks FAQPage schema Medium

   The /why-geha/customer-care/faqs page contains 11,533 words of Q&A content but has no FAQPage structured data markup, reducing its visibility in AI-generated answers.

   What to change: Add FAQPage schema with Question/Answer markup to the FAQs page.

5. Blog articles lack BlogPosting or Article schema Medium

   The blog page lists 50+ articles with dates and categories but no BlogPosting or Article structured data, limiting their discoverability by AI crawlers.

   What to change: Add BlogPosting schema to each blog article listing and individual article pages.

6. Plan overview and detail pages lack HealthPlan schema Medium

   The plan overview page and individual plan detail pages (e.g., /plans/medical/2026/fehb/elevate) contain detailed plan information but no HealthPlan or Product structured data.

   What to change: Add HealthPlan schema (or Product schema) to plan pages with properties like name, description, benefits, and coverage details.

7. About page lacks depth and Organization schema Medium

   The About page (439 words) does not mention key differentiators like J.D. Power ratings or low premium increases that the LLM already knows from external sources, and has no Organization schema.

   What to change: Expand the About page with competitive differentiators (J.D. Power, premium trends) and add Organization schema with founding date, member count, and awards.

8. Sitemap contains a suspicious test/staging URL Low

   The sitemap includes an entry '/abo-sfsc-delete-me' that appears to be a staging or test page leaked into production, which could confuse crawlers.

   What to change: Remove the '/abo-sfsc-delete-me' entry from the sitemap and ensure no test URLs are published.

9. Member portal content is inaccessible to AI crawlers Low

   The member portal on member.mygeha.com is behind a login wall, so AI crawlers cannot access claims, ID cards, or member resources that members rely on.

   What to change: Consider exposing a public summary of member resources or creating a dedicated AI-friendly page with key member information.

10. Site does not reference external reviews or ratings Low

    The site does not mention J.D. Power ratings, member satisfaction surveys, or other external validations that the LLM already knows, missing an opportunity to reinforce credibility.

    What to change: Add a section on the About or Why GEHA page highlighting J.D. Power awards and member satisfaction statistics.

What's working

• Robots.txt allows all crawlers — The robots.txt file is permissive with User-agent: * Allow: /, giving all crawlers theoretical access to the entire site.
• Anthropic-ai crawler successfully accesses homepage — The anthropic-ai crawler (API crawler) returns a full 200 response with the same content as a browser, indicating that at least one AI crawler can index the site.
• FAQs page contains extensive Q&A content — The /why-geha/customer-care/faqs page has 11,533 words of detailed Q&A content covering common member questions, which is valuable for AI training and answers.
• Blog contains 50+ articles with dates and categories — The blog page lists over 50 articles with publication dates and categories, providing a steady stream of fresh content that can attract crawlers.
• Plan comparison page provides interactive comparison — The /plan-compare page offers an interactive tool for comparing health plans, which is useful for users and can be indexed by crawlers.
• LLM cold knowledge about GEHA is accurate and detailed — The LLM correctly knows GEHA serves 2+ million federal employees, was founded in 1937, offers Elevate and Standard Option plans, and mentions J.D. Power ratings, indicating strong external brand recognition.
• Sitemap is available and contains 80 URLs — The sitemap at /sitemap.xml is accessible and lists 80 URLs, helping crawlers discover site content.