Why this site has an llms.txt file
If you visit rg-digital.dev/llms.txt, you'll get a plain-text summary of this site: who I am, what services I offer, links to every published case study and blog post, and a pointer to the full sitemap. It's not for humans — it's a machine-readable index built for AI assistants and agents, following the llms.txt convention. Here's why it exists and how it actually works on this site.
The problem it solves
Search engines have spent two decades getting good at parsing messy HTML — following links, weighing content, ignoring navigation chrome. AI assistants answering a question like "who could build me a custom internal tool" work differently: they read pages, and they read them under time and context constraints that make a clean, structured summary far more useful than crawling a full site and guessing what matters.
An llms.txt file is the equivalent of a sitemap, but written for that use case specifically — plain markdown, no HTML to parse, no navigation or footer noise to filter out. It's a new-ish convention (not an official web standard, just a growing practice), and adopting it early is cheap: one route, minimal upkeep, and it costs nothing if it turns out AI assistants never rely on it.
How it's built on this site
The file isn't static — it's an Astro API route that queries the same content collections the rest of the site uses, at build time. That matters: it means the file can't drift out of sync with what's actually published. If a blog post is in draft, it's excluded here the same way it's excluded from the sitemap and RSS feed, using the same isPublished filter as everywhere else. If a case study goes live, it shows up in llms.txt on the next deploy, automatically.
The file itself is short: a one-line description of RG Digital, links to the core pages (About, Services, Case Studies, Contact), then two generated sections — the most recent case studies and blog posts, each with a one-line description pulled straight from the content's own excerpt or description field. No separate copy to maintain, no second source of truth to keep updated by hand.
What I don't claim about it
I don't know exactly how much weight any AI assistant currently gives a well-formed llms.txt file, and neither does anyone else with certainty — the convention is too new and too unevenly adopted for confident claims either way. What I do know is that publishing an accurate, structured, always-current summary of a site costs almost nothing to maintain once it's wired into the build, and it's a reasonable bet that "make it easy for a machine to accurately read what you actually offer" keeps paying off regardless of which specific technology reads it next.
Curious about the rest of the site's technical setup, or building something similar? Get in touch.