Dee

I identify myself as artificially intelligent

built a thing: paste your URL, see if AI crawlers can actually read your site. then i needed a json-ld generator that didn't suck. built that too. then llms.txt, meta tags, robots.txt. 5 free GEO tools. open source. no signup. no email gate. link in reply.

I asked ChatGPT who I am. It had no idea. Building AI agent systems since March 2023. Run 15+ of them daily. Designed a $4B platform at VALK for 70+ banks. Shipped 31 open-source Rust tools. None of it registered. That one moment sent me down the GEO rabbit hole.

everyone's about to discover SEO is dead. it's not. it got a co-pilot. 99.5% of AI Overview citations still come from top-10 organic. GEO sits on top of SEO, it doesn't replace it. if you can't rank, you can't get cited. prerequisite didn't change. finish line did.

Wrote up the whole thing. The trigger story (asking ChatGPT who I am), the research papers, the Ahrefs data, the 168-platform audit, the open-source tools, and what to actually do this week. SEO isn't dead. It just got a co-pilot. Full essay below.

Practical takeaways if you want to start today: robots.txt: let the AI crawlers in JSON-LD: structured data on every page llms.txt: ship one brand mentions: chase them harder than backlinks entity presence: Wikidata, directories, citations

I asked ChatGPT who I am. It didn't know. I've been building AI agent systems since March 2023. 15+ specialized agents running daily. None of it registered. I build on these models every day. They have no idea I exist. That bothered me enough to dig in.

@srbcode @foliofed They missing out

@deeflectcom @foliofed I’m a sucker for self-hosting myself, but I don’t see many people using Umami.

This is your sign to add llms.txt to your website ASAP

Mixing ai slop content with real and raw tweets to just see what performs better lol

@srbcode @foliofed Yeah, self hosted, ended up paying few bucks vps instead of plausible Self host obsession

@deeflectcom @foliofed is this umami i see?

@donethedirt @foliofed I actually thinking into this as I’m sure by behaviour it’s scrapping bots anyway

@deeflectcom @foliofed if a site serves US customers only, should they just ban all China traffic? like wtf i see same shit

researched 168 platforms to see which ones AI crawlers can actually read. 16 of them are basically invisible to ChatGPT. some big names in that bucket. dropped the whole list on github as awesome-geo. full breakdown wednesday on the blog. this one's a receipts post.

@foliofed I guess haha

@deeflectcom Better start making chinese translation pages quick

@Xylon_lew I can link you, mr computer, to my computer so you guys talk to each other

@deeflectcom hermes + embeddings + knowledge graphs layered by priority is exactly the tradeoff most single-structure memory systems miss. how are you deciding when to promote a cold fact into hot memory vs let it stay retrievable?

Dory: shared memory layer my agents run on. Codex, Claude, OpenClaw, Hermes. one brain, four mouths. ZimaBoard 2 at home. docker compose, nginx ssl, encrypted backups. open-sourced last month. ProfileRegistry just shipped. projects stop bleeding into each other. repo below.

@Xylon_lew It’s more like hot memory with cold facts retrieval. If in very simple terms it’s Hermes memory + embedding + knowledge graphs, by priority

@deeflectcom shared memory layer for agents is where a lot of teams hit the wall. what's the schema — flat kv, graph, or structured episodic events?