I ran 5 websites. Published 58 articles. To AI, I didn't exist.

I've been busy.

Five live sites. Fifty-eight articles. Three SaaS products. Weeks of after-hours and weekends pushing things out into the world. LinkedIn posts. A portfolio worth pointing at.

I figured I'd built a nice little layer of authority.

I was wrong.

I put myself to the test

One day I asked a handful of AI tools what they knew about me.

Fresh sessions. No memory. Pure research.

The answers made me cringe.

Most of my sites referred to me as plain "Ben". No last name. One had my name wrong entirely. To any AI asking "who built this?", I was a ghost with a portfolio.

The "I'll fix it later" trap

Here's the part that stung.

I knew about this stuff. Person schema. Author tags. rel="me". It was on the list.

Right after launching SerpDelta.
Right after the next LayerView release.
Right after the next push on itbroke.dev.

Never got there.

So I ran a proper audit. Five sites, five layers of identity signals. Every layer came back empty.

No Person schema on any site.
No <meta name="author">. Anywhere.
No rel="me" in the <head>.
llms.txt files that didn't mention my full name.
Footers that said "Launch With Ben" but never "Ben Peetermans". A brand name, not a person.
My /about page returned a 404.

Months of content. Months of promotion. A machine-readable identity worth exactly zero.

Why this matters

Old-school SEO is about content. LLM search is about entities.

When someone asks an AI "who built X?" the answer doesn't come from scanning your homepage and guessing. It comes from structured signals. Person schema. Organization schema with a founder property. sameAs links that connect your LinkedIn to your products. Author meta on every article.

If those signals aren't there, the AI doesn't guess. It just… doesn't answer.

Or it answers wrong. Like one of mine did.

What actually fixes it

Two hours. No paid tools.

1. Person schema on your main site. One JSON-LD block. [@type](/type): Person, your full name, a sameAs array listing your LinkedIn, your socials, and every product you've built. The [@id](/id) is the important part. It's the anchor every other page can reference. One identity, referenced everywhere.

2. Organization schema with a founder property. On your product sites, link the business back to the person. Don't create disconnected "Organization" entries with no human attached. founder points at your Person [@id](/id). That's the glue.

3. Author meta tags on every page. <meta name="author" content="Your Full Name">. Not a nickname. Not "Ben". Not "The Team". Your actual name.

4. Fix your bylines. If your articles just say "by Ben", they're useless. "By Ben Peetermans" is how machines recognize a person. First name alone is a ghost.

5. Footer attribution. Every page. "Built by [Your Full Name]". I had "Launch With Ben". A brand name isn't a person. Fixed.

6. A /about page that doesn't 404. Mine did. This is the canonical "who is this person" destination. If it doesn't exist, every footer link you add points at nothing.

7. rel="me" links in the <head>. <link rel="me" href="https://linkedin.com/in/you">. Repeat for each profile and each product. This is how the decentralized web says "these all belong to the same person". Free. Skipped by almost everyone.

8. llms.txt with your full name. If you have one, it probably just describes what your site does. Add your name. Link to an About page or press kit. This is the file AI crawlers read for context.

9. humans.txt. Old-school, still works. One file, your name, your URL. 30 seconds.

10. Cross-site links, both directions. Your products should link back to you. Your About page should link to your products. The entity graph is a graph. It needs edges.

How to check if it worked

Three free tests:

Google's Rich Results Test. Paste a URL. See if Person and Organization parse cleanly.
Schema.org Validator. Deeper. Catches missing [@id](/id) references.
Ask ChatGPT or Perplexity: "Who built [your-site.com]?" If the answer is confident and correct, you did it right. If it hedges or gets it wrong, something's still missing.

I caught two silent bugs running these. Breadcrumb schema producing doubled URLs. article:published_time outputting a Unix timestamp instead of ISO 8601. Both "looked right". Neither was.

The part I wish I'd heard two years ago

This isn't SEO in the gaming sense. No tricks. No hacks.

It's just telling the truth in a format machines can read.

You're building in public. Running multiple properties. Shipping on nights and weekends. Posting on LinkedIn. And wondering why AI tools never mention you when they talk about your space.

This is probably why.

Two hours. Ten signals. A checklist your junior self could follow.

Now when an AI gets asked who built any of my sites, the answer isn't "plain Ben". Isn't the wrong person. Isn't silence.

It's me.

If you try the "who built [your-site.com]?" test on your own stuff, I'm curious what you see. Drop the result in the comments. I'll do a few live.