I Built This Site to Be Found by AI. Here’s Exactly What I Did.

I teach businesses how to get found by AI. That would be a little hollow if this site weren’t built the same way — so it is. Below is the actual setup: what I did to GetDiscoverable.io, why each piece matters to a machine, and how you’d do the same. Where I use a technical term, I define it on the spot.

First, the honest part. This site is new. As I write this, it hasn’t earned an AI citation yet, and it isn’t ranking on the hard search terms — because none of this is instant. Search engines send out automated readers (called crawlers) to visit your pages and file them away (called indexing); then rankings and trust build over weeks. So treat this as “here’s the work, done right,” not a victory lap. I’ll report back on what actually gets cited.

Now the moves.

1. I wrote pages whose headline is the exact question people ask

Not clever titles — literal questions, like “How do I make my business discoverable on AI?” AI assistants tend to quote the page that most directly asks and answers a question, so the headline is the strategy. Each post leads with a short, plain answer up top, because the machine grabs the first clean one it finds.

2. Everything important is real text

No key information trapped inside images or PDFs. If you can’t highlight it with your cursor, a machine can barely read it — so every fact and every answer lives on the page as selectable text.

3. I added structured data — the “hidden labels”

There’s a small chunk of code you can add to a page that never shows on screen. Its only job is to tell a machine what each thing on the page is: “this is the author,” “this is a service,” “this is a question and its answer.” It’s called structured data, and it’s like the tag sewn into a shirt collar — invisible to the shopper, but it spells out exactly what everything is. My question-and-answer pages use a specific kind of it (called FAQ markup) so those Q&As are eligible to be quoted directly.

4. I wrote a brief for the AI, in a file called llms.txt

It’s a plain-text file — literally named llms.txt — that sits on the site and gives AI systems a short summary of who I am, what the site is, and the facts they’re welcome to repeat. It’s written for the machines, not for visitors. Most sites don’t have one, so it’s a small, low-effort way to hand AI a clean summary instead of making it assemble one.

5. I made sure the AI readers are allowed in

Every site has a small text file called robots.txt that tells those automated readers what they’re allowed to look at. Mine doesn’t just allow Google — it names and welcomes the AI readers specifically: GPTBot (ChatGPT’s reader), PerplexityBot, ClaudeBot, Google’s Gemini crawler, and Apple’s. Plenty of sites accidentally block these and never realize it. If the readers can’t get in, you can never be recommended — so I made sure the door is open.

6. I go by one name, everywhere

It’s “GetDiscoverable.io” on every page, in the page’s hidden details, and anywhere I appear elsewhere online. Consistency teaches the machine I’m one trusted thing, not three faint ones.

7. I linked my own pages to each other on purpose

A central questions page that points to the deeper posts; posts that point back to the playbook and the case study. Those internal links — links from one of my pages to another — help both readers and the machines understand which pages matter and how they connect.

8. The boring plumbing is all in place

None of this is glamorous. All of it removes friction:

• A sitemap — a simple list of every page on the site — so search engines can find them all.
• An RSS feed — a machine-readable list of my latest posts that apps and services can subscribe to.
• A canonical URL on every page — the single official web address for that page, so search engines don’t get confused when the same page can be reached more than one way (for example, with or without “www,” or with a campaign tag on the end).
• Social-share cards — the little preview (image, title, and blurb) that appears when someone shares a link.
• A fast, lightweight build with no heavy software behind it, so a crawler isn’t slowed down.

And off the site: I answered real questions where AI looks

AI assistants lean on a few places they already trust — so I answered genuine questions on them: LinkedIn, plus honest, useful replies on Reddit and Quora threads about this topic. Not spam drops — real answers that happen to sit on the surfaces Google’s AI answers cite most.

What I’m deliberately not doing

Buying ads. This is the same playbook that got my other project, COLAClear, named in Google’s AI Overview (the AI-written answer at the top of Google) with zero paid media. Getting found by AI rewards being readable and trustworthy, not outspending anyone.

How I’ll know it worked

I’ll watch Google Search Console (Google’s free tool that shows which searches your pages appear in) for the first signs of these queries, re-run the searches myself every couple of weeks to see if my pages or my Reddit and Quora answers get surfaced, and check whether ChatGPT and Perplexity start naming the site when asked. When they do, I’ll write the follow-up — receipts included.

Zillah Bahar is the founder of COLAClear and writes about getting found by AI at GetDiscoverable.io.