mirakl.com — AI Site Grade

Analysis

The blog is a client-rendered JS shell that delivers zero article content to any AI crawler

Mirakl.com is hosted on Netlify (A record 75.2.60.5) and built on Next.js. Every AI crawler tested — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Bytespider, Applebot-Extended, anthropic-ai — receives a 200 status with identical byte size to a browser visit on the homepage and all static pages. No UA-based blocking exists. The robots.txt is a bare-bones Allow: / with no AI-bot-specific directives. The llms.txt exists and is well-structured, listing products, use cases, and resources with descriptions and URLs — a strong signal.

Cold-Knowledge Gap

The LLM's prior knowledge describes Mirakl as a "SaaS platform for enterprises to launch and manage online marketplaces" founded in 2012 in Paris, with products including "Mirakl Marketplace Platform, Mirakl Connect, and Mirakl PIM." The cold model knows about the $300M Series E in 2021 at a $3.5B valuation and mentions Carrefour and Best Buy Canada as clients. The site itself has moved aggressively beyond this positioning. The homepage now leads with "Mirakl Nexus: The Agentic Commerce Brain" and "Agentic Activation" — a product line for making product catalogs discoverable and purchasable on LLMs like Microsoft Copilot. The cold model knows nothing about Mirakl Nexus, Agentic Activation, or the Stripe partnership for AI-platform selling. The site's entire top-level narrative has pivoted to agentic commerce, but the model's prior is stuck on the 2021-era marketplace-platform story.

Schema Posture

Zero JSON-LD structured data was found on any page examined — the homepage, the Nexus page, the about page, the technology page, the marketplace platform page, the agentic activation page, and the pricing page all return empty schema.jsonld arrays. A company selling an "Agentic Activation" product that enriches product catalogs for LLM discoverability does not itself use any schema markup (Organization, WebSite, BreadcrumbList, FAQPage, Product) on its own site. The Agentic Activation page does contain an FAQ section with 9 questions and answers rendered in plain HTML — a prime candidate for FAQPage schema that is entirely absent.

Blog Content Invisible to AI Crawlers

The /news-and-blog page returns a Next.js client-rendered shell — only 8 words of visible text ("All news & blog Type: All Topic: All") are extractable from a plain GET. The blog posts listed on the homepage (e.g., "Why most retailers aren't ready for agentic commerce" dated May 28, 2026) are loaded dynamically and are inaccessible to AI crawlers. The guessed URL for that article returns a 404. The sitemap contains 339 URLs across three locales (en-US, fr-FR, ja-JP), but the blog content — the site's most timely, AI-relevant material — is effectively invisible to every crawler.

External Signals

DNS TXT records confirm domain verification for OpenAI (openai-domain-verification=dv-WPlTgDvKM92SX3YDnyzqLTbQ), Anthropic (anthropic-domain-verification-vxzwnz), and Cursor — indicating active relationships with AI platforms. The site references a Stripe partnership for selling on AI platforms, a $11.2B GMV figure for 2024, and 100% uptime since the start of 2024. Customer logos include Carrefour, Best Buy Canada, Macy's, La Redoute, Showroomprive, Conrad, Toyota Material Handling, and Madewell.

Findings

1. Blog delivers zero article content to AI crawlers High

   The /news-and-blog page is a Next.js client-rendered shell returning only 8 words of visible text to plain GET requests. Blog posts listed on the homepage are loaded dynamically and inaccessible to AI crawlers. The guessed URL for a featured article returns a 404.

   What to change: Implement server-side rendering or static generation for blog content so that AI crawlers can extract full article text. Ensure blog post URLs are included in the sitemap and return 200 with complete content.

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

   No JSON-LD structured data was found on any examined page, including the homepage, Nexus page, about page, technology page, marketplace platform page, agentic activation page, and pricing page. The Agentic Activation page contains an FAQ section with 9 questions and answers in plain HTML, a prime candidate for FAQPage schema that is entirely absent.

   What to change: Add Organization, WebSite, BreadcrumbList, and FAQPage JSON-LD structured data to relevant pages. Prioritize FAQPage schema on the Agentic Activation page.

3. Cold LLM knowledge is stuck on 2021-era marketplace positioning High

   The LLM's prior knowledge describes Mirakl as a marketplace platform founded in 2012, with a $3.5B valuation from 2021. The site has pivoted to agentic commerce with products like Mirakl Nexus and Agentic Activation, but the cold model knows nothing about these. This gap means AI-generated summaries will misrepresent the company's current offerings.

   What to change: Publish authoritative content (press releases, blog posts, case studies) about Mirakl Nexus and Agentic Activation on the site, and ensure it is crawlable and indexed. Consider submitting to AI training data sources.

4. Blog URLs missing from sitemap High

   The sitemap contains 80 URLs across three locales, but no blog post URLs are included. The blog content is the most timely and AI-relevant material, yet it is effectively invisible to crawlers.

   What to change: Add all blog post URLs to the sitemap and ensure they return 200 with full content.

5. No AI-bot-specific directives in robots.txt Low

   The robots.txt is a bare-bones Allow: / with no directives for AI bots like GPTBot, ClaudeBot, or Google-Extended. While not blocking, it misses an opportunity to guide crawlers to important content.

   What to change: Add specific directives for AI bots, such as allowing access to blog and product pages, and consider using crawl-delay to manage load.

6. Weak external signals from reviews and analyst reports Medium

   Web searches for Mirakl reviews on G2, Forrester, and Gartner returned zero results. This limits third-party validation that AI models might surface.

   What to change: Encourage customers to leave reviews on G2, Gartner Peer Insights, and other platforms. Publish case studies and analyst reports on the site.

What's working

• Well-structured llms.txt file — The llms.txt file exists and is well-structured, listing products, use cases, and resources with descriptions and URLs. This provides a strong signal to AI crawlers about key site content.
• No user-agent blocking for AI crawlers — All tested AI crawlers receive 200 status with full content on static pages. No UA-based blocking exists, ensuring broad access.
• Domain verified with OpenAI, Anthropic, and Cursor — DNS TXT records confirm domain verification for OpenAI, Anthropic, and Cursor, indicating active relationships with AI platforms that can improve visibility.
• Dedicated Agentic Activation product page with FAQ — The Agentic Activation page explains how the product makes product catalogs discoverable on LLMs, and includes an FAQ section with 9 questions and answers. This is directly relevant to AI visibility.
• Sitemap covers three locales — The sitemap includes URLs for en-US, fr-FR, and ja-JP locales, supporting international AI visibility.