heur.co.uk — AI Site Grade

Analysis

Heur's llms.txt with MCP endpoint is ahead of most agencies, but the rest of the AI-visibility stack is bare-minimum — and the cold-knowledge gap is severe, with LLMs hallucinating a Shopify-only Gymshark agency that does not match what the site actually sells.

Crawler Access

All major AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, anthropic-ai, Bytespider, Applebot-Extended — receive a 200 status with real HTML content from the Wix-hosted site. No UA-based blocking exists. The robots.txt at heur.co.uk/robots.txt contains a User-agent: * Allow: / rule but names zero AI bots explicitly. The llms.txt at heur.co.uk/llms.txt is present and unusually sophisticated: it exposes a Wix Model Context Protocol (MCP) endpoint (/_api/mcp) enabling AI agents to query business details, search the site, and even call Wix APIs directly. This is a rare and forward-leaning implementation for an agency site.

Cold-Knowledge Gap

A frontier LLM queried cold describes Heur as a "UK-based Shopify Plus development agency" whose notable clients include Gymshark and who specializes in headless commerce and custom Shopify builds. The actual site contradicts this on every point. Heur's homepage and service pages describe a platform-agnostic fractional ecommerce agency offering consultancy, managed ecommerce, and performance marketing. The Drake's case study explicitly describes a Magento-to-Shopify migration — not a Shopify-native build. Gymshark does not appear anywhere on the site among the 8+ named clients (Drake's, Union Coffee, Tangle Teezer, King & McGaw, H&M, NICCE, Far Afield, TOG Knives). The LLM hallucinated a positioning that does not exist.

Schema Posture

The homepage carries only a bare WebSite schema with name and URL — no Organization, Service, FAQPage, or LocalBusiness schema anywhere on the site. The blog post about ChatGPT and ecommerce discovery has a proper BlogPosting schema with author, dates, and headline. All other key pages — consultancy, management, performance marketing, case studies, about — have zero JSON-LD schema. Given the site's explicit claims of being "award-winning" and "multi-award winning," no Award property or Review schema is present.

External Signals

Web searches for "Heur ecommerce agency" return zero indexed results from any third-party source — no press coverage, no review sites, no Reddit threads, no industry publication mentions. The only external backlinks found on the site are to LinkedIn and Instagram profiles. The DNS TXT records show a Klaviyo verification token and Google Search Console verification, confirming active marketing tools, but the brand has no detectable off-domain footprint that AI engines could cite.

Content Architecture

Every page on the site duplicates the same global navigation block (three H1s, ~20 H2s) as visible DOM content before reaching page-specific body text. This means AI crawlers parsing the page encounter ~300 words of boilerplate navigation before the unique content. The blog is active (7 posts, latest from July 2025) and includes a post specifically about AI-driven ecommerce discovery — but the post is a guest contribution from an external SEO agency (Blink SEO), not original Heur thought leadership. The case studies page lists 8 clients but individual case study pages contain only ImageObject schema, no structured result data.

Findings

1. LLMs hallucinate Heur as a Shopify-only Gymshark agency High

   Frontier LLMs describe Heur as a Shopify Plus agency with Gymshark as a client, but the site is platform-agnostic and does not mention Gymshark. This mismatch undermines AI-generated citations.

   What to change: Add explicit platform-agnostic messaging and client lists to structured data and llms.txt to correct LLM misconceptions.

2. No Organization or LocalBusiness schema on any page High

   The homepage only has a bare WebSite schema. Key pages like consultancy, management, and about lack Organization, Service, or LocalBusiness schema, reducing AI understanding of the business.

   What to change: Add Organization, LocalBusiness, and Service schema to all relevant pages, including award and review properties.

3. No detectable third-party mentions or backlinks High

   Web searches for Heur return zero results from press, reviews, Reddit, or industry publications. The brand has no off-domain footprint for AI engines to cite.

   What to change: Build external backlinks through guest posts, case study syndication, and industry directory listings.

4. Duplicate navigation boilerplate dominates page content Medium

   Every page repeats the same global navigation block (~300 words) before unique body text, diluting the signal-to-noise ratio for AI crawlers.

   What to change: Reduce navigation HTML weight or use JavaScript to defer rendering of navigation for crawlers.

5. Case study pages lack structured result data Medium

   Individual case study pages contain only ImageObject schema, no schema for results, metrics, or client testimonials, limiting AI extraction of success stories.

   What to change: Add schema markup for case studies including metrics, client name, and results.

6. Robots.txt does not explicitly name AI crawlers Low

   The robots.txt has a generic Allow rule but does not explicitly allow or disallow AI bots like GPTBot or ClaudeBot, leaving behavior to defaults.

   What to change: Add explicit directives for major AI crawlers to ensure optimal crawling.

7. AI-focused blog post is guest-contributed, not original Low

   The blog post about ChatGPT and ecommerce discovery is written by an external SEO agency, reducing Heur's perceived authority on AI topics.

   What to change: Publish original thought leadership on AI and ecommerce to build authority.

What's working

• llms.txt with MCP endpoint enables AI agent queries — Heur's llms.txt exposes a Wix MCP endpoint, allowing AI agents to query business details and APIs directly, a rare and forward-leaning implementation.
• All major AI crawlers receive 200 HTML responses — GPTBot, ClaudeBot, and other AI crawlers get full HTML content with no blocking, ensuring content is indexable.
• Blog post has correct BlogPosting schema — The ChatGPT blog post includes proper BlogPosting schema with author, dates, and headline, aiding AI extraction.
• Active blog with recent posts on relevant topics — The blog has 7 posts, latest from July 2025, including AI-related content, demonstrating ongoing content production.
• Sitemap.xml is present and contains 29 URLs — A valid sitemap helps crawlers discover all pages efficiently.