rnid.org.uk — AI Site Grade

Analysis

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

RNID: AI crawlers get full access but find a site that contradicts its own cold-knowledge reputation

The most striking finding is that every major AI crawler (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User) receives a full 200 with 345KB of content — yet the cold LLM knowledge about RNID is stuck on a 2023 advertising controversy and the old "Action on Hearing Loss" name, while the actual site is a polished, modern WordPress deployment with no trace of that controversy anywhere.

Crawler Access

The robots.txt is a bare Yoast-generated file with no AI-bot-specific rules — just a catch-all User-agent: * disallowing only file-type extensions (.pdf, .xls, .docx). No GPTBot, ClaudeBot, or Google-Extended directives exist. compare_bot_access confirmed that Bytespider is the sole blocked crawler (403 from Cloudflare), while all other AI bots get the same 345KB HTML payload as a browser. The site runs on Cloudflare (CDN) + WP Engine (hosting) with nginx serving 404s. No llms.txt exists (404). The sitemap is a Yoast index with 6 sub-sitemaps and ~700+ URLs, all well-structured.

Cold-Knowledge Gap

The LLM's prior knowledge of RNID is stale and partially inaccurate. It recalls the 2011–2020 "Action on Hearing Loss" trading name, mentions a "Louder" campaign and an online hearing check tool, and references a 2023 advertising controversy that "some deaf communities felt misrepresented." The actual site makes no mention of any controversy. The homepage headline is "Hearing loss can strike suddenly" — a spring appeal, not a campaign. The site describes itself as supporting 18 million people (not 12 million as the LLM states). The LLM knows nothing about the charity's research funding (£31 million, 472 projects), its BSL Act campaigning, or its current "It Does Matter" campaign. The gap between what AI models would say cold and what the site actually communicates is wide.

Schema Posture

Every page carries rich JSON-LD schema via Yoast SEO: WebPage, Organization, WebSite, BreadcrumbList, ImageObject, and SearchAction. The Organization block correctly lists sameAs links to Facebook, X, Instagram, LinkedIn, YouTube, and Wikipedia. However, no FAQPage, HowTo, Article, MedicalWebPage, or HealthTopicContent schema is used anywhere — surprising for a health-information charity with pages on hearing loss, tinnitus, and ear health. The information pages are structured as plain text with headings, not as structured answer formats. No FAQ sections, comparison tables, or definition lists were detected on any page sampled.

External Signals

The Wikipedia entry is comprehensive and well-maintained (last edit April 2026), covering the charity's 1911 founding, the 2011 rebrand to Action on Hearing Loss, the 2020 reversion to RNID, and recent financial challenges including a "going concern" warning from PwC in 2018. The site itself does not reference these financial difficulties. The DNS records show Microsoft 365 (Outlook) for mail, Zoho verification, and Facebook/Google site verification — a standard charity tech stack. No negative press or controversy surfaced in web searches, suggesting the 2023 advertising issue the LLM recalled may have been minor or misattributed.

Surprising Details

The homepage datePublished in schema reads "2026-05-08" — a future date, suggesting a content scheduling artifact or a Yoast misconfiguration. The site uses duplicate heading content (the same H3 blocks appear twice in the homepage markup, likely a carousel or grid component rendering both visible and hidden states). The research page claims "over £31 million" and "472 projects" in the body text, but the meta description still says "over £28 million" and "443 projects" — a stale meta description that AI crawlers will index as the summary. The "About us" page states the charity serves "18 million people" while the Wikipedia page and LLM knowledge both say "12 million" — a discrepancy that will confuse AI engines synthesizing information from multiple sources.

Findings

1. Research page meta description is stale Medium

   The research page body text claims 'over £31 million' and '472 projects', but the meta description still says 'over £28 million' and '443 projects'. AI crawlers often use meta descriptions as page summaries, so this mismatch undermines accuracy.

   What to change: Update the meta description on the research page to match the current body text figures.

2. Homepage schema has a future datePublished Medium

   The homepage JSON-LD schema includes a datePublished value of '2026-05-08', which is in the future. This may confuse AI crawlers and search engines about the page's freshness.

   What to change: Correct the datePublished in the homepage schema to the actual publication or last-modified date.

3. Health information pages lack specialized schema Medium

   Pages about hearing loss, tinnitus, and ear health use only generic WebPage schema. No MedicalWebPage, HealthTopicContent, FAQPage, or HowTo schema is present, which limits their visibility in AI-generated health answers.

   What to change: Add MedicalWebPage or HealthTopicContent schema to health-related pages, and consider FAQPage for common questions.

4. LLM prior knowledge is outdated and inaccurate High

   LLMs recall RNID's old trading name 'Action on Hearing Loss', a 2023 advertising controversy, and a 12 million figure, while the site uses 'RNID' exclusively, mentions no controversy, and claims 18 million people. This gap means AI models may misrepresent the charity.

   What to change: Publish an llms.txt file and ensure key facts (name, mission, impact numbers) are consistently stated across the site and in structured data.

5. No llms.txt file published Medium

   The site returns a 404 for /llms.txt. This file would allow RNID to provide authoritative facts to AI crawlers, helping correct the cold-knowledge gap.

   What to change: Create an llms.txt file with key facts about RNID's mission, impact numbers, and current campaigns.

6. Homepage has duplicate heading content Low

   The homepage markup contains identical H3 blocks appearing twice, likely from a carousel or grid component rendering both visible and hidden states. This can confuse crawlers about content uniqueness.

   What to change: Use CSS or JavaScript to hide duplicate elements instead of duplicating them in the HTML.

7. Audience number discrepancy between site and Wikipedia Medium

   The site states it supports '18 million people', while Wikipedia and LLM knowledge say '12 million'. This inconsistency will confuse AI synthesizing multiple sources.

   What to change: Align the audience number across the site and consider updating Wikipedia to reflect the current figure.

8. Bytespider crawler is blocked Low

   Bytespider (ByteDance's crawler) receives a 403 from Cloudflare, while all other major AI crawlers are allowed. This may limit visibility in ByteDance's AI products.

   What to change: If RNID wants visibility in ByteDance's AI ecosystem, allow Bytespider access.

What's working

• All major AI crawlers are allowed — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and others receive full 200 responses with rich HTML content. No AI-specific disallow rules exist in robots.txt.
• Every page has comprehensive JSON-LD schema — Yoast SEO injects WebPage, Organization, WebSite, BreadcrumbList, and SearchAction schema on all pages. Organization block includes correct sameAs links to social media and Wikipedia.
• Sitemap is well-structured with sub-sitemaps — The Yoast sitemap index contains 6 sub-sitemaps covering ~700+ URLs, ensuring all pages are discoverable by crawlers.
• Wikipedia entry is comprehensive and up-to-date — The Wikipedia page covers RNID's history, rebranding, financial challenges, and campaigns, and was last edited in April 2026. This provides a strong external signal for AI knowledge bases.
• Organization schema is consistent across pages — The Organization schema block is identical on all sampled pages, providing consistent name, logo, and social links.