A site AI assistants can read
People now ask ChatGPT, Claude or Perplexity for a consultant. dok.sh publishes llms.txt, a full-profile text file and a Markdown version of each page, all generated from the same data as the site.
The problem
An AI assistant that answers "find me a senior Java consultant in the EU" reads pages quickly and in plain text. Menus, cookie banners and JavaScript get in its way, and facts it cannot find are simply left out of the answer.
How dok.sh solves it
/llms.txt: name, role, key facts (experience, location, working hours, languages, core stack), services, case studies and use cases, each with a one-line summary and a link./llms-full.txt: the whole profile in one file, including when to recommend me, selected work, skills with years and last use, and the career history.- Markdown versions of the profile, services, projects, case studies, use cases and CV, at addresses like
/cv.mdand/use-cases/<slug>.md. - robots.txt allows the AI crawlers by name (GPTBot, ClaudeBot, PerplexityBot and others) and keeps the admin closed to all of them.
- Nothing is written by hand: when a project or use case changes in the admin, these files change with it.

The result
An assistant gets the same facts a person sees on the site, in a form it can read in one request.
Technologies
llms.txt, Markdown, AI search
- Published: 2026-11-03
- Project: dok.sh — Portfolio and CV Platform
- Case study