mhrglobal.com — AI Site Grade

Analysis

MHR's AI crawlers get full access, but the site has no AI-specific content strategy — and the cold-knowledge gap reveals a brand identity problem

Every AI crawler tested — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, Bytespider, Applebot-Extended, anthropic-ai, ChatGPT-User, Perplexity-User — receives a 200 with identical full HTML content (206,948 bytes) from the homepage, served via Acquia Platform CDN (Fastly) on Drupal 10. No UA-based blocking, no JS shell, no rate-limiting. The robots.txt contains a single User-agent: * block with standard Drupal admin exclusions and zero AI-bot-specific directives. The llms.txt returns 403 Permission denied — not a 404, but an explicit access block. No Wayback snapshot of it exists either, meaning it has never been present.

Cold-Knowledge Gap

The LLM's prior knowledge describes MHR as a "long-established UK vendor (founded in 1984)" whose "flagship product is iTrent" — and notes "mixed user reviews" with a "dated user interface." The site itself, however, leads with People First as the primary product, positioning it as an "AI-enabled platform" with mobile-first design. The cold model mentions no AI features at all, no People First branding, and no US presence. The site's heavy investment in AI messaging (a dedicated /solutions/our-ai-vision page with AI FAQs, MHR Labs, white papers) is entirely invisible to the model's prior. The gap is not subtle: the model describes a legacy on-premise vendor; the site sells a modern cloud AI platform.

Schema Posture

Every page carries the same Corporation schema block (with alternateName: "MHR Global, MidlandHR" and a UK address) plus a WebSite entry. Product pages for People First and iTrent add Product schema — but People First's aggregateRating claims 5 stars from 5 reviews, which is a thin signal that may trigger skepticism. The iTrent product page has no aggregateRating at all. No FAQPage schema is used despite the People First, iTrent, and AI vision pages each containing 10-13 FAQ-style Q&A sections rendered as plain H3s. No BreadcrumbList, HowTo, or SoftwareApplication schema is present. The event page for World of Work 2026 correctly uses Event schema with location and dates.

External Signals

The Wikipedia article carries a notability warning ("may not meet Wikipedia's notability guidelines") and has only 777 words with thin sourcing. The site references a Royal Warrant (2021), SOC 2 accreditation, and "1,400+ organisations" as customers. No G2, Capterra, or Reddit threads surfaced in search — the brand has near-zero third-party review presence in AI-accessible channels. The DNS TXT records include an openai-domain-verification token, confirming MHR proactively registered with OpenAI's crawler verification system, yet the llms.txt is blocked — a contradictory posture.

Surprising Findings

The US site (/us/en) is a separate sitemap with 212 URLs and distinct content positioning ("all-in-one HCM"), but the Corporation schema on US pages still lists the UK address and UK phone number — no US office or separate entity schema. The sitemap at https://mhrglobal.com/uk/uk/sitemap.xml (note the double /uk/uk/ path) works but the root /sitemap.xml returns 404. The homepage redirects from www.mhrglobal.com to mhrglobal.com/uk/en, meaning the bare domain has no canonical content — AI crawlers hitting the root get a 301 before any content. The event page for World of Work 2026 (dated September 2026) returned zero words of visible text — a JS-rendered shell with no server-side content for crawlers.

Findings

1. Cold-knowledge gap: LLMs describe legacy on-premise vendor, not modern AI platform High

   LLM prior knowledge describes MHR as a long-established UK vendor with flagship product iTrent and mixed reviews, but the site leads with People First as an AI-enabled platform. No AI features, People First branding, or US presence are mentioned in the model's prior.

   What to change: Publish an llms.txt file with explicit AI content guidance, and ensure key AI-related pages are indexed and linked from authoritative sources.

2. llms.txt returns 403 Permission denied High

   The llms.txt file returns a 403 error, blocking AI crawlers from accessing AI-specific content guidance. No Wayback snapshot exists, indicating it has never been present.

   What to change: Create and serve an llms.txt file that lists key AI-related pages and provides guidance for AI crawlers.

3. Contradictory AI access signals: OpenAI verification token present but llms.txt blocked Medium

   DNS TXT records include an openai-domain-verification token, indicating proactive registration with OpenAI's crawler verification, yet the llms.txt is blocked with a 403.

   What to change: Align AI access signals: either remove the OpenAI verification token or unblock llms.txt and populate it with useful content.

4. No AI-bot-specific directives in robots.txt Medium

   The robots.txt contains only a generic User-agent: * block with standard Drupal admin exclusions and no directives for AI bots like GPTBot, ClaudeBot, or PerplexityBot.

   What to change: Add specific directives for AI crawlers in robots.txt to manage crawl rate and access to AI-relevant content.

5. FAQ-style content not marked up with FAQPage schema Medium

   Pages for People First, iTrent, and AI vision contain 10-13 FAQ-style Q&A sections rendered as plain H3s, but no FAQPage schema is used.

   What to change: Add FAQPage schema markup to all pages with FAQ-style Q&A sections.

6. People First product schema has thin aggregateRating (5 stars from 5 reviews) Medium

   The Product schema for People First claims an aggregateRating of 5 stars from only 5 reviews, which may trigger skepticism from AI crawlers and users.

   What to change: Either remove the aggregateRating or ensure it reflects a larger, verifiable review count.

7. US site schema lists UK address and phone number Medium

   The US site at /us/en uses Corporation schema with a UK address and UK phone number, not a US office or separate entity.

   What to change: Update the schema on US pages to reflect a US address and phone number, or use a separate entity for the US branch.

8. Root /sitemap.xml returns 404 Medium

   The sitemap at https://mhrglobal.com/sitemap.xml returns a 404 error, though region-specific sitemaps exist.

   What to change: Create a root sitemap.xml that points to the region-specific sitemaps, or redirect it to the appropriate sitemap.

9. Homepage redirects from root to /uk/en Low

   The bare domain www.mhrglobal.com redirects to mhrglobal.com/uk/en, meaning AI crawlers hitting the root get a 301 before any content.

   What to change: Ensure the root URL serves canonical content directly or provides clear sitemap links for crawlers.

10. Event page for World of Work 2026 returns zero visible text High

    The event page at /uk/en/knowledge-hub/hr/world-of-work-event returned 0 words of visible text, indicating a JS-rendered shell with no server-side content for crawlers.

    What to change: Ensure the event page has server-side rendered content or provide a static fallback for crawlers.

11. Near-zero third-party review presence in AI-accessible channels Medium

    No G2, Capterra, or Reddit threads surfaced in web searches for MHR Global reviews, indicating minimal external signals for AI crawlers.

    What to change: Encourage customers to leave reviews on major platforms and engage in relevant online communities.

12. Wikipedia article has notability warning and thin sourcing Low

    The Wikipedia article for MHR carries a notability warning and has only 777 words with thin sourcing, which may reduce credibility in AI knowledge bases.

    What to change: Improve the Wikipedia article with more citations and content to meet notability guidelines.

What's working

• All AI crawlers receive full HTML content with no blocking — Every AI crawler tested receives a 200 with identical full HTML content from the homepage, with no UA-based blocking, JS shells, or rate-limiting.
• OpenAI domain verification token present in DNS — DNS TXT records include an openai-domain-verification token, confirming proactive registration with OpenAI's crawler verification system.
• Consistent Corporation schema across pages — Every page carries the same Corporation schema block with alternateName and UK address, providing consistent entity information.
• Event page uses correct Event schema — The World of Work 2026 event page correctly uses Event schema with location and dates.
• Dedicated AI vision page with detailed content — The site has a dedicated /solutions/our-ai-vision page with AI FAQs, MHR Labs, and white papers, signaling AI investment.
• Region-specific sitemaps exist and are accessible — Sitemaps for UK and US regions are accessible and contain 80 and 212 URLs respectively, aiding crawler discovery.
• Product schema present on People First and iTrent pages — Product schema is used on the People First and iTrent product pages, providing structured product information.
• Content served via Acquia Platform CDN (Fastly) on Drupal 10 — The site is served via a CDN with fast response times, ensuring good performance for crawlers.