safeguardglobal.com — AI Site Grade

Analysis

Safeguard Global's /campaign/ robots.txt block is aspirational, not enforced — every AI crawler tested returns a full 200 with identical byte size to browser baseline on campaign pages, despite 27 explicit Disallow rules.

Crawler Access

All 11 AI crawlers tested (GPTBot, ClaudeBot, anthropic-ai, Google-Extended, PerplexityBot, OAI-SearchBot, ChatGPT-User, Applebot-Extended, Bytespider, Perplexity-User, Browser) receive identical 200 responses with identical byte size (383,353 bytes) on the homepage and campaign pages alike. The robots.txt contains 27 AI-specific Disallow: /campaign/ rules with a comment explaining these pages are "noindex'd at the page level" — but the block is entirely honor-based. No HTTP 403, no Cloudflare challenge, no UA-based gating. Campaign pages like /campaign/employer-of-record-services/ serve full HTML to GPTBot and every other bot. The site runs on AWS CloudFront + AmazonS3 with HSTS and X-Content-Type-Options: nosniff but no CSP or X-Frame-Options.

llms.txt and Content Architecture

The site has an exemplary llms.txt — 23KB of structured content covering solutions, use cases, country resources, and explicit guidance to AI models (e.g., "Do not infer support terms, SLAs, or legal commitments beyond what is explicitly stated"). This is among the most thorough llms.txt files observed in the EOR space. The sitemap-index.xml contains 2,652 URLs across country pages, blog posts, and solution pages. The blog publishes multiple times weekly with content dated into May 2026, indicating aggressive forward-publishing or future-dated content.

Cold-Knowledge Gap

The LLM cold-knowledge snapshot describes Safeguard Global as a "stable, mid-tier player" founded in 2006 in Austin, TX, competing with Deel and Remote. The site itself says founded in 2008 (a 2-year discrepancy). The LLM knows nothing about the 2025 sale of the enterprise payroll division to Deel, the Intelligent Workforce AI tool launched April 2026, the NelsonHall 2025 Leader designation, or the "Human when it matters" brand positioning. The LLM also has no awareness of the ChatSG generative AI tool announced July 2023. The gap between cold knowledge and live site content is substantial — the site has evolved significantly in the last 12-18 months while the model's prior is frozen at a pre-2024 snapshot.

Schema and Answer Signals

The homepage carries a single Organization schema with knowsAbout listing 9 service categories, foundingDate: 2008, and areaServed: Worldwide. The EOR solution page adds FAQPage schema with 4 questions (including "What is an EOR?" and "EOR vs PEO") and a Service schema with hasOfferCatalog. The Intelligent Workforce page also carries FAQPage schema with 6 detailed questions. The Deel comparison page uses tabular comparison language (industry tenure, hidden fees, compliance) but lacks a Comparison schema type. No Product or SoftwareApplication schema is present for the platform itself.

External Signals

The newsroom shows consistent press coverage through mid-2026, including Inc., HR Dive, and TechR Series. The sale of the enterprise payroll division to Deel (March 2025) is a major strategic event that the cold LLM does not reference. The site carries Anthropic domain verification and Apple domain verification TXT records, confirming proactive AI-ecosystem registration. G2 reviews are linked in the footer but web searches returned zero organic review results — the G2 presence is not generating discoverable search snippets.

Findings

1. Robots.txt blocks on campaign pages are not enforced High

   The robots.txt contains 27 AI-specific Disallow rules for /campaign/ pages, but all 11 tested AI crawlers receive a full 200 response with identical byte size (383,353 bytes) as a browser baseline. No HTTP 403, Cloudflare challenge, or UA-based gating is applied.

   What to change: Remove the Disallow rules for /campaign/ from robots.txt, or implement server-side enforcement (e.g., 403 response or Cloudflare challenge) for AI bots that should be blocked.

2. LLM cold knowledge is significantly outdated High

   The LLM snapshot describes Safeguard Global as a mid-tier player founded in 2006, but the site states 2008. The model lacks awareness of the 2025 sale of the enterprise payroll division to Deel, the Intelligent Workforce AI tool, the NelsonHall 2025 Leader designation, the ChatSG generative AI tool, and the 'Human when it matters' brand positioning.

   What to change: Increase external signals (press releases, analyst reports, social media activity) to help LLMs update their knowledge. Consider submitting structured data updates to knowledge panels.

3. Comparison page lacks structured schema Medium

   The Deel comparison page uses tabular comparison language but does not include a Comparison or Product schema type, reducing its ability to appear in rich results or be understood by AI crawlers.

   What to change: Add Comparison or Product schema to comparison pages to enable rich snippets and improve AI comprehension.

4. Platform lacks SoftwareApplication schema Medium

   No Product or SoftwareApplication schema is present for the platform itself, which limits its ability to be indexed as a software product by AI crawlers and search engines.

   What to change: Add SoftwareApplication schema to the platform and product pages to improve AI visibility and rich result eligibility.

5. G2 reviews not generating organic search snippets Medium

   Web searches for Safeguard Global G2 reviews returned zero results, indicating the G2 presence is not discoverable via search engines, reducing external signal strength.

   What to change: Encourage more G2 reviews and ensure the G2 profile is linked with proper schema markup to improve discoverability.

6. Blog contains future-dated content Low

   The blog publishes content dated into May 2026, indicating aggressive forward-publishing or future-dating. This may confuse AI crawlers about content freshness and accuracy.

   What to change: Ensure blog post dates reflect actual publication dates to maintain credibility with AI crawlers and search engines.

7. Missing CSP and X-Frame-Options headers Low

   The site lacks Content-Security-Policy and X-Frame-Options headers, which are not directly related to AI visibility but indicate a gap in security posture that could affect trust signals.

   What to change: Add Content-Security-Policy and X-Frame-Options headers to improve security posture.

What's working

• Exemplary llms.txt with 23KB of structured content — The site has a thorough llms.txt file covering solutions, use cases, country resources, and explicit guidance to AI models, among the best observed in the EOR space.
• FAQPage schema on EOR and Intelligent Workforce pages — The EOR solution page and Intelligent Workforce page include FAQPage schema with detailed questions, improving eligibility for rich results and AI comprehension.
• Anthropic and Apple domain verification TXT records — The site carries Anthropic and Apple domain verification TXT records, confirming proactive registration in AI ecosystems.
• Consistent press coverage through mid-2026 — The newsroom shows regular press mentions from Inc., HR Dive, and TechR Series, providing external signals that can help LLMs update knowledge.
• Organization schema with knowsAbout and areaServed — The homepage carries a single Organization schema listing 9 service categories, founding date, and worldwide coverage, providing clear entity information to AI crawlers.
• Sitemap with 2,652 URLs covering country pages and blog — The sitemap-index.xml contains 2,652 URLs, ensuring comprehensive coverage of site content for crawlers.