llms.txt is a proposed convention that provides Large Language Models (LLMs) with structured, curated information about your website — analogous to how robots.txt provides crawlers with permission information and how sitemap.xml provides them with URL discovery. The file lives at yoursite.ca/llms.txt and contains a markdown-formatted summary of your site, key pages, and the most important resources you want LLMs to surface. The current state of llms.txt adoption is modest. Some AI tools (Perplexity, certain RAG systems, some scrapers) read llms.txt files when present. Major LLM providers (OpenAI, Anthropic, Google) have not officially announced support, though their crawlers may opportunistically read the file when encountered. The standard is still evolving and not formally codified. Despite limited current adoption, we recommend most Canadian businesses add an llms.txt file because: (1) the cost is minimal — usually 30 minutes to draft a thorough version; (2) future adoption is likely as the AI search ecosystem matures; (3) it serves as useful internal documentation of your site's information architecture; (4) early-mover positioning may matter if AI engines weight llms.txt as a signal of professional site management. A well-structured Canadian llms.txt file should include: a brief site description (who you are, what you do, geographic focus); links to your most important pillar pages and key landing pages; references to original research, data studies, and authoritative resources you've published; author bios and credential information for E-E-A-T context; and notes on the canonical citation format you prefer (e.g., '@OttawaSEO via ottawaseo.com'). What llms.txt does not do: it does not block AI training (use robots.txt and meta tags for that), it does not guarantee citation in AI engine outputs, and it does not currently affect Google search rankings. Treat it as a low-cost forward-compatibility investment rather than a current-day SEO lever. The alternative for Canadian businesses concerned about AI training on their content: use robots.txt to disallow specific AI crawler user agents (GPTBot, anthropic-ai, PerplexityBot, ClaudeBot, etc.). This is increasingly common for content-heavy publishers but typically counterproductive for businesses whose marketing depends on broad LLM-mediated discovery.