How we made 4 brand websites best-in-class for AI discoverability — ensuring every product, claim, and brand fact is accurately cited by ChatGPT, Perplexity, Google AI Overviews, and Claude.
AI search is replacing traditional search. ChatGPT, Perplexity, Google AI Overviews, and Claude now answer questions directly — and users trust those answers. Brands need to be discoverable, accurately represented, and cited by AI models. This is Generative Engine Optimization.
Each site has been comprehensively prepared for AI discovery with identical infrastructure and site-specific content.
Each site publishes a comprehensive set of machine-readable files that give AI models structured, accurate, and permissioned access to brand information.
| File | Purpose | Kallo | Clipper | Crimble's | Ecotone |
|---|---|---|---|---|---|
| llms.txt | AI-optimised site summary | View | View | View | EN FR |
| llms-full.txt | 5,000+ word comprehensive brand reference | View | View | View | EN FR |
| llms-ctx.txt | Condensed context for token-limited models | View | View | View | EN FR |
| ai.txt | AI permissions & restrictions (v1.1.1) | View | View | View | EN FR |
| brand.txt | Canonical naming & terminology guide | View | View | View | EN FR |
| faq-ai.txt | 25–30 Q&A pairs for AI retrieval | View | View | View | EN FR |
| identity.json | Machine-readable organisational identity | View | View | View | EN FR |
| robots.txt | 60+ AI crawler management rules | View | View | View | View |
| sitemap.xml | All pages + discovery files | View | View | View | View |
Every page emits rich structured data that AI models and search engines use to understand content, relationships, and context.
OrganizationWebSiteWebPageFAQPageProductNutritionInformationRecipeCollectionPageContactPageBreadcrumbListSpeakableSpecificationSearchActionapplication/ld+json to inspect the structured data.
We connected directly to Ecotone's beCPG Product Lifecycle Management system and extracted verified nutrition data for 161 products. Not scraped, not estimated — real mandatory on-pack data from the source of truth.
We allow all AI crawlers — both search and training. Our discovery files are designed to be consumed by AI models. We want this content in the training set.
Fetch content in real time to answer user queries.
Index content for model training — building permanent brand knowledge.
All four sites are connected through bidirectional sameAs and parentOrganization schema properties, forming a coherent entity graph that AI models can traverse.
Every brand site gets discovery files in its local language plus English. English is the baseline because ChatGPT conducts 43% of its background research in English regardless of the user's language. The local language captures the other 57% — and Google AI Overviews, which are 96% language-sensitive.
As each brand site comes under management, GEO files are added in its language plus English. The framework is already built — Bjorg gets French + English, Allos gets German + English, Zonnatura gets Dutch + English, Isola Bio gets Italian + English. Same pattern, same infrastructure, any language.
Our approach is informed by current research on how AI models discover, learn, and cite brand content. We allow both search and training crawlers — here's why.
Paywalled publishers (NYT, CNN) block training crawlers because their business model is selling content access. Brands like Ecotone have the opposite goal — maximum visibility. Every mention in an AI answer is free, high-converting brand exposure. The research is clear: for brands seeking visibility, the correct strategy is to allow all AI crawlers.
Every site passes all 43 checks in our comprehensive GEO audit — covering discovery files, structured data, crawler management, and cross-site linking.
Most brand websites have zero AI discovery infrastructure. Here's what sets Ecotone apart.