What I learned rebuilding GIA.net for modern discovery because humans aren't the only ones reading it anymore.
I started noticing something strange while rebuilding GIA.net.
Humans understood the site almost immediately. AI systems kept seeing old information, missing context, or misunderstanding what the site actually was. In my case, they kept confusing me with the famous diamond people.
That sent me down a rabbit hole. Modern websites aren't just being read by people anymore. They're being interpreted by AI systems, search engines, summaries, assistants, and crawlers that all process information differently.
AI systems don't experience your website the way humans do. They aren't emotionally reacting to your branding or your beautiful animations. They're trying to piece together meaning from structure, text, metadata, and relationships.
That means clarity matters.
While rebuilding this site, I started experimenting with a few things designed to help AI systems better understand what GIA actually is and does.
llms.txt AI-readable site summaryllms-full.txt Consolidated full-text markdown payloadindex.md markdown version of key contextsitemap.xml structured guidance for crawlersrobots.txt clear crawler instructionsPlus: cleaner semantic page structure, consistent identity signals, and tighter messaging across the board.
None of this is magic. It's just making the site easier to interpret.
The web has changed a lot since the before times when I was hand-coding in Pico directly on the server. People still use search engines, but they're increasingly discovering businesses, products, and ideas through AI-generated summaries and assistants. And often not by choice - the browser is just feeding them AI results first.
If AI systems misunderstand your site, they may describe it incorrectly, surface outdated info, miss what makes you unique or fail to reference you at all.
Helping machines understand your site clearly is becoming part of modern discoverability.
I told a few people what I was working on. A couple were a little confused because "Isn't that just SEO?". No. One shrugged it off and said "It's too early to worry about that". Also no.
The funny part is I didn't hand-code most of it. Those days are far behind me. I used AI to help generate and refine the structure, summaries, markdown files, and discovery layer and help me understand the why.
Here's the process I used:
prompts what I actually askedworkflows order of operationsstarter files ready to customizedeployment notes where things goNot because the files themselves are complicated, but because figuring out what matters can feel overwhelming when you're starting from zero.
Everything I used to make GIA.net legible to AI systems. Free, obviously.
