Proven in production — LLMO · Large Language Model Optimization
The setup runs in production on blackbook.dk — llms.txt, numerous markdown twins and sitemap.md, read by AI crawlers every day. We advise on how your site gets read — not just found.
Services
Learned, not just found.
Your brand is known before anyone asks.
Learned, not just found
SEO and GEO are about being found and cited. LLMO is about living in the model's knowledge – so you are the answer before anyone searches.
Deterministic, no guessing
The AI gets the text served directly instead of guessing its way through markup and scripts. Fewer misquotes, stronger answers.
Open to AI crawlers
A robots policy that invites AI crawlers in instead of shutting them out. Measurable in the logs: GPTBot, ClaudeBot, PerplexityBot and Meta's crawlers read the site every day.
One dataset, every format
HTML, JSON-LD and markdown are generated from the same source. Humans, search engines and AI models read the same content – never three versions drifting apart.
llms.txt
One index for AI
A curated index at the root of the site telling AI models exactly what exists and where to find it. The standard behind AI-readable websites – blackbook.dk/llms.txt is listed in llms.txt directories and read by agents daily.
index.md
Every page as plain text
Every page has a markdown twin: same content, zero noise. Where the HTML version carries design and the knowledge graph, the .md layer is built for agents that want the text directly – deterministic, no scraping guesswork.
sitemap.md
A sitemap for agents
Search engines get sitemap.xml – agents get sitemap.md. The complete list of markdown twins in a single request, so an AI can read the entire site without rendering a single pixel.
robots.txt
An open door for AI
Many sites block AI crawlers without knowing it. Our robots policy invites them in — GPTBot, ClaudeBot, PerplexityBot and Meta's crawlers — and points them to the right layers instead of shutting the door.
canonical + json-ld
One identity, one preferred link
A knowledge graph with stable IDs for person, organisation and brand — plus a canonical and preferred citation URL on every page. The AI is never in doubt about who the sender is or which link to cite.
The identity — llms.txt
One file at the root tells the AI who the sender is: entities with stable IDs, the preferred citation URL and the path to the rest of the machine-readable layer.
The twin — index.md
Every page has a markdown twin. The frontmatter ties the twin to its canonical page, language and last-updated date — and the entity IDs point back to the sender. Same content, zero noise.
Cases the AI can retell
Inside the twin, each case is written as structured text with a link to its source. That lets the AI retell the work accurately — and point back to the right page.
Discovery — the way in
The AI has to find everything on its own. The door is open, the signals are explicit, and the signposts point onwards: robots.txt welcomes crawlers and links straight to sitemap.md and llms.txt — the whole machine-readable layer is discoverable from a single file.
The case
Live
blackbook.dk—thewholesiteisthecase
llms.txt, numerous .md twins and sitemap.md run in production on this very site. AI crawlers fetch thousands of pages a day, and Blackbook is cited in answers from ChatGPT, Perplexity and Claude. The proof isn't a client case — it's the site you're looking at.
See the sitemap for AI agents