freshfields.com — AI Site Grade

Analysis

Excellent — the RSS feed is rich with actual content. Now I have a comprehensive picture. Let me write the audit.

Freshfields.com: A JS-Rendered Shell Behind Open Bot Gates

The site's core pages — about-us, capabilities, our-thinking, locations, campaign articles, blog pages — all return zero words of visible text to every AI crawler tested, despite every major AI bot (GPTBot, ClaudeBot, Google-Extended, PerplexityBot, OAI-SearchBot, ChatGPT-User, anthropic-ai, Applebot-Extended) receiving a 200 status with no UA-based blocking.

Crawler Access

The robots.txt at freshfields.com/robots.txt contains no AI-bot directives whatsoever. No GPTBot, no ClaudeBot, no Google-Extended, no PerplexityBot — only a catch-all User-agent: * with a few disallow rules (blogs, CMS paths, noindex assets) and explicit Allow: / for SemrushBot. The llms.txt returns a 404 — the site has no AI-friendly content map. Bytespider is the sole bot blocked, returning 403 from Cloudflare. All other AI crawlers pass through Cloudflare unhindered but receive a Next.js client-side rendering shell: ~250-760 KB of HTML containing CSS bundles, JS chunk references, and a loading spinner GIF, but no article body text, no paragraph content, no substantive visible text on any sub-page tested.

Content & Schema Posture

The homepage is the only page that server-renders meaningful text (437 words, including featured insights and navigation). Every other page — about-us, capabilities, our-thinking, locations, campaign landing pages, blog pages, podcast pages, article pages — returns 0 words of extracted text. The site runs on Next.js with client-side rendering for all sub-pages. Zero JSON-LD schema was found on any page tested — no Organization, WebSite, LegalService, Article, FAQPage, or BreadcrumbList schema. The site has no FAQ, comparison, or table answer-format signals. The RSS feed at /en/rss/news is the only machine-readable content source, containing full-text news articles (9,683+ words) that the HTML pages fail to deliver.

Cold-Knowledge Gap

The LLM knows Freshfields as a Magic Circle firm headquartered in London, strong in cross-border M&A, antitrust, and international arbitration, serving multinational corporations and governments. It recalls the firm's role in UK privatizations, its pro bono program, and recent reputational scrutiny over fossil fuel client work. The site itself never mentions "Magic Circle" anywhere on the homepage or about-us page — a significant positioning gap. The site's tagline ("Law never stops moving forward – neither do we") and "Fresh thinking for a changing world" are generic brand messaging that does not map to the firm's specific market identity as a Magic Circle leader. The cold knowledge about AI tool adoption for document review is also absent from the site's visible content.

External Signals

The DNS TXT records reveal an anthropic-domain-verification token, confirming Freshfields has proactively verified its domain with Anthropic for Claude integration. Other verification tokens include Google, Apple, Deepl, Canva, Zoom, Atlassian, DocuSign, and multiple analytics/email platforms (Salesforce Marketing Cloud, Pardot, Mimecast). The site links to X (Twitter), LinkedIn, and Xing but has no visible social proof, client testimonials, or third-party review aggregation on-page. The sitemap contains ~6,181 URLs across English, German, Japanese, and Chinese (traditional) — a massive content corpus that is almost entirely invisible to AI crawlers due to JS rendering.

Findings

1. Core pages render as empty JS shells to AI crawlers High

   Every sub-page tested (about-us, capabilities, our-thinking, locations, campaign articles, blog pages) returns 0 words of visible text to GPTBot, ClaudeBot, Google-Extended, PerplexityBot, and others. The HTML contains only CSS, JS chunks, and a loading spinner, with no article body or paragraph content.

   What to change: Implement server-side rendering (SSR) or static generation for all sub-pages so that AI crawlers receive full HTML content. Alternatively, use dynamic rendering to serve pre-rendered content to known bot user agents.

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

   No JSON-LD schema was found on the homepage or any sub-page. Missing Organization, WebSite, LegalService, Article, FAQPage, and BreadcrumbList schemas that would help AI models understand the site's structure and content.

   What to change: Add JSON-LD structured data for Organization, WebSite, LegalService, Article, and BreadcrumbList on relevant pages. Use schema.org types appropriate for a law firm.

3. Robots.txt has no AI-bot directives Medium

   The robots.txt file contains no rules for GPTBot, ClaudeBot, Google-Extended, PerplexityBot, or other AI crawlers. Only a catch-all User-agent: * with a few disallow rules and an explicit Allow for SemrushBot. This leaves AI crawlers unguided but they still receive empty shells.

   What to change: Add explicit directives for AI crawlers in robots.txt, either allowing them to access pre-rendered content or disallowing if desired. Also consider adding a Crawl-delay directive.

4. llms.txt returns 404 Medium

   The site has no llms.txt file, which is a recommended way to provide an AI-friendly content map. This means AI crawlers have no guidance on which pages to prioritize or how to interpret the site structure.

   What to change: Create an llms.txt file listing key pages (about, capabilities, thinking, locations) with brief descriptions to help AI crawlers discover and understand the site's content.

5. Magic Circle positioning absent from site content Medium

   The site never mentions 'Magic Circle' on the homepage or about-us page, despite the firm being widely known as a Magic Circle member. The generic tagline 'Law never stops moving forward' does not convey the firm's elite market position.

   What to change: Incorporate 'Magic Circle' and other distinctive positioning language into the homepage and about-us page to align with the firm's market identity and improve AI knowledge retrieval.

6. RSS feed is the only machine-readable content source Medium

   The RSS feed at /en/rss/news contains full-text news articles (9,683+ words) but the HTML pages fail to deliver this content to crawlers. The feed is a valuable asset that is not being leveraged for AI visibility.

   What to change: Ensure that the content available in the RSS feed is also server-rendered on the corresponding HTML pages. Consider using the RSS feed as a source for an AI-friendly content API or sitemap.

7. No FAQ, comparison, or table answer-format content Medium

   The site lacks FAQ pages, comparison tables, or other structured answer formats that AI models often use to generate direct answers. This limits the site's ability to appear in featured snippets or AI-generated answers.

   What to change: Create FAQ pages for common legal questions and practice areas. Use FAQPage schema markup to help AI models extract structured answers.

8. No client testimonials or third-party reviews on-page Low

   The site does not display client testimonials, case studies, or third-party review aggregation. This reduces trust signals for AI models and users alike.

   What to change: Add a testimonials section or link to third-party review sites (e.g., Chambers, Legal 500) on the homepage or about-us page.

9. Bytespider bot blocked by Cloudflare Low

   The Bytespider crawler receives a 403 response from Cloudflare, preventing it from accessing the site. While this may be intentional, it could limit visibility on Baidu's AI platforms.

   What to change: If the site wants visibility on Chinese AI platforms, consider allowing Bytespider access. Otherwise, no action needed.

10. Sitemap index file not referenced in robots.txt Low

    The sitemap index at /sitemapindex.xml exists but is not declared in robots.txt. This may delay discovery of the 6,181 URLs across multiple languages.

    What to change: Add a Sitemap directive in robots.txt pointing to the sitemap index URL.

What's working

• All major AI crawlers allowed access — No AI bots are blocked by robots.txt or Cloudflare (except Bytespider). GPTBot, ClaudeBot, Google-Extended, PerplexityBot, and others receive 200 responses, indicating the site intends to be crawled.
• Anthropic domain verification token present — DNS TXT records include an anthropic-domain-verification token, confirming Freshfields has proactively verified its domain with Anthropic for Claude integration. This is a strong signal of AI readiness.
• Homepage server-renders meaningful text — The homepage returns 437 words of visible text, including featured insights and navigation, making it the only page that provides content to AI crawlers without JavaScript execution.
• RSS feed provides full-text news articles — The RSS feed at /en/rss/news contains 9,683+ words of full-text news articles, offering a machine-readable content source that can be used by AI crawlers if properly linked.
• Sitemap covers 6,181 URLs across multiple languages — The sitemap index includes URLs in English, German, Japanese, and Chinese (traditional), indicating a large content corpus that, if rendered properly, could provide substantial AI visibility.
• Cloudflare provides DDoS and bot protection — The site uses Cloudflare, which offers security and performance benefits. The cf-ray headers indicate active protection, and Bytespider is blocked, suggesting some bot management is in place.
• Social media links to X, LinkedIn, and Xing — The site includes links to its social media profiles, which can help AI models discover external signals and verify the brand's online presence.
• Cold knowledge aligns with firm's actual reputation — The LLM's prior knowledge correctly identifies Freshfields as a Magic Circle firm strong in M&A, antitrust, and arbitration, which matches the firm's actual market position. This provides a foundation for AI visibility if the site content aligns.