sozodesign.co.uk — AI Site Grade

Analysis

The cold-knowledge gap is the single biggest finding: LLMs describe SOZO as a small local Cheltenham web design shop serving SMEs, while the actual site positions itself as a 25-year-old global B Corp digital growth agency with dedicated AI agent building and AI search optimisation services.

Crawler Access

All 11 AI crawlers tested (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Bytespider, Applebot-Extended, anthropic-ai, Perplexity-User, and a browser baseline) receive a 200 status with identical 500KB payload from Cloudflare. No UA-based blocking exists. The robots.txt is minimal — a single User-agent: * rule disallowing only /wp/wp-admin/ — with zero AI-bot-specific directives. DNS records confirm both openai-domain-verification and anthropic-domain-verification TXT tokens are present, indicating proactive verification with both major LLM providers. The site runs on Kinsta (Laravel 12/Acorn) behind Cloudflare, with strong security headers (HSTS preload, CSP, X-Frame-Options).

llms.txt and Content Architecture

The site has a fully populated llms.txt (54KB, generated by Rank Math SEO) listing dozens of blog posts with descriptions, plus a sitemap reference. This is an advanced signal — few agencies at this scale have deployed it. The sitemap index contains 9 sub-sitemaps (post, page, career, portfolio, video, webinar, ebook, category, local) with 229+ URLs. The blog is heavily weighted toward AI topics: AI search optimisation, agentic commerce, entity disambiguation, AIO vs SEO, and Shopify AI visibility. The homepage itself contains an FAQ section with an explicit question "How can we improve our visibility in AI chatbots?" — the agency is literally eating its own dog food.

Cold-Knowledge Gap

A frontier LLM queried cold describes SOZO as a "digital agency based in Cheltenham, UK, specializing in web design, branding, and digital marketing services" serving "small to medium-sized businesses, local enterprises, and startups" with "custom WordPress websites, e-commerce solutions, and brand identity packages." This is stale by several years. The actual site describes a 25-year-old agency with offices in Cheltenham, Cape Town, and Mumbai, a certified B Corp, building custom AI agents, offering dedicated AI search optimisation (AIO/GEO) services, and counting clients like The Whisky Exchange, Trojan Energy, and Smart Home Sounds — brands with measurable revenue growth figures (+86%, +143%, +253%) displayed prominently. The model knows nothing about the AI agent building service, the B Corp status, the global offices, or the AIO positioning.

Schema Posture

JSON-LD is present on every page fetched but is structurally weak. The Organization schema lacks sameAs links (no LinkedIn, YouTube, or social profiles in the schema despite the site having them), has no foundingDate, numberOfEmployees, award, or review markup despite the site displaying Google 5-star reviews and client testimonials. The WebSite schema includes SearchAction but the potentialAction target URL uses a generic {search_term_string} placeholder. Blog posts use BlogPosting type but the author entity is a generic Gravatar hash named "sozo_V2s4GHcr" — not a real person — which undermines EEAT signals. No FAQPage schema is used on pages that contain FAQ content (homepage, ecommerce page, AI agents page).

External Signals

Web searches returned zero indexed results for the brand name, the domain, or common queries about SOZO — an unusual vacuum that suggests either very recent domain changes, low external citation footprint, or DuckDuckGo limitations. The site itself references client logos (The Whisky Exchange, Trojan Energy, Smart Home Sounds, Beet It, Nutrition X, Just Fabrics) and displays specific growth metrics, but no independent third-party verification of these case studies was found in search. The LinkedIn company page and YouTube channel exist (linked in footer) but are not surfaced in schema.

Findings

1. LLMs describe SOZO as a small local web design shop, missing global B Corp AI agency positioning High

   A frontier LLM queried cold describes SOZO as a small Cheltenham web design shop serving SMEs, while the actual site positions itself as a 25-year-old global B Corp digital growth agency with AI agent building and AI search optimisation services. The model knows nothing about the AI services, B Corp status, global offices, or major clients.

   What to change: Publish a dedicated 'AI visibility' page or update the homepage and about page with explicit statements about AI services, B Corp status, and global offices. Ensure this content is crawlable and included in llms.txt with high-quality descriptions.

2. Organization schema lacks sameAs, foundingDate, awards, and reviews Medium

   The JSON-LD Organization schema on every page omits sameAs links (LinkedIn, YouTube), foundingDate, numberOfEmployees, award, and review markup, despite the site displaying Google reviews and client testimonials. This limits rich snippets and entity understanding by AI crawlers.

   What to change: Add sameAs URLs for LinkedIn, YouTube, and other social profiles; include foundingDate, numberOfEmployees, award, and review properties in the Organization schema.

3. Blog author entity uses generic Gravatar hash instead of real person name Medium

   Blog posts use BlogPosting schema with an author entity that has a generic Gravatar hash ('sozo_V2s4GHcr') as the name, not a real person. This undermines EEAT signals and reduces trustworthiness in AI-generated summaries.

   What to change: Replace the Gravatar hash with the actual author name and add author profile markup (e.g., sameAs, jobTitle) to strengthen EEAT.

4. FAQ content on multiple pages lacks FAQPage schema markup Medium

   Pages containing FAQ sections (homepage, ecommerce page, AI agents page) do not use FAQPage schema, missing an opportunity for rich results and direct answer extraction by AI crawlers.

   What to change: Add FAQPage schema markup to all pages with FAQ content, using the Question and Answer properties.

5. No indexed search results found for brand or domain queries High

   Web searches for the brand name, domain, and common queries returned zero indexed results, indicating very low external citation footprint or potential indexing issues. This limits the site's visibility in AI training data and search engines.

   What to change: Investigate indexing status via Google Search Console, build backlinks from reputable sources, and ensure the site is properly indexed. Consider a PR or content syndication strategy to increase external citations.

6. Robots.txt has no AI-bot-specific directives Low

   The robots.txt file contains only a single User-agent: * rule disallowing /wp/wp-admin/, with no specific directives for AI crawlers like GPTBot or ClaudeBot. While not blocking, this misses an opportunity to guide crawlers to important pages.

   What to change: Add specific directives for AI crawlers, e.g., allow all for GPTBot and ClaudeBot, and consider a separate sitemap for AI-optimized content.

7. WebSite SearchAction uses generic placeholder URL Low

   The WebSite schema includes a SearchAction with a potentialAction target URL using a generic {search_term_string} placeholder, which may not work correctly for AI crawlers and reduces the quality of the schema.

   What to change: Update the SearchAction target URL to the actual search endpoint, e.g., https://sozodesign.co.uk/?s={search_term_string}.

8. Client case studies lack independent third-party verification Medium

   The site displays client logos and growth metrics (+86%, +143%, +253%) but no independent third-party sources or links to case study pages were found in search results, reducing credibility for AI models that prioritize verifiable claims.

   What to change: Publish detailed case study pages with client testimonials, links to client sites, and third-party references. Ensure these pages are indexed and linked from the main site.

What's working

• Fully populated llms.txt file with blog post descriptions and sitemap reference — The site has a 54KB llms.txt generated by Rank Math SEO, listing dozens of blog posts with descriptions and a sitemap reference. This is an advanced signal that helps AI crawlers discover and understand content.
• All 11 tested AI crawlers receive 200 status with no UA-based blocking — All major AI crawlers (GPTBot, ClaudeBot, PerplexityBot, etc.) receive a 200 status and identical payload, with no user-agent blocking. The robots.txt only disallows /wp/wp-admin/.
• DNS TXT records include openai and anthropic domain verification tokens — The DNS records contain both openai-domain-verification and anthropic-domain-verification TXT tokens, indicating proactive verification with both major LLM providers, which can improve crawling and indexing.
• Sitemap index with 9 sub-sitemaps covering 229+ URLs — The sitemap index contains sub-sitemaps for posts, pages, careers, portfolio, video, webinar, ebook, category, and local, ensuring comprehensive coverage of site content for crawlers.
• Blog heavily weighted toward AI topics including AIO, agentic commerce, and entity disambiguation — The blog contains numerous posts on AI search optimisation, agentic commerce, entity disambiguation, and AIO vs SEO, positioning the site as an authority on AI visibility topics.
• Homepage FAQ explicitly asks about improving visibility in AI chatbots — The homepage includes an FAQ section with the question 'How can we improve our visibility in AI chatbots?', demonstrating the agency's focus on AI visibility and providing relevant content for AI crawlers.
• Site uses strong security headers including HSTS preload and CSP — The site employs HSTS preload, Content Security Policy, and X-Frame-Options headers, which contribute to trust signals for AI crawlers and users.