We built Ghost to help companies be visible to AI agents.
So, naturally, the first thing we did was point our scanner at tryghost.tech (our own landing page). We expected a perfect 100/100 score. After all, we wrote the rules, right?
We scored a 32.
The "Cobbler's Children" Problem
It turns out, being an expert in AI optimization doesn't automatically mean your code reflects it. In the rush to build features, we verified every user flow except the one for AI agents.
Here is exactly what we failed at:
1. Missing llms.txt (-40 Points)
We preach that llms.txt is the "Front Door" for AI agents. It's a simple text file that tells Claude, ChatGPT, and Perplexity exactly what your site does.
We forgot to add one.
2. No Structured Data (-10 Points)
Our landing page looked great to humans, but to an AI, it was just a soup of <div> tags. We had no JSON-LD schema to explain "This is a SaaS product" or "Here is our pricing".
3. Strict Robots.txt (-10 Points)
Irony of ironies: We were blocking the very bots we wanted to embrace. Our default robots.txt from a template was set to Disallow: /, effectively telling the AI world to go away.
How We Fixed It (In 10 Minutes)
Because Ghost tells you exactly what is wrong, the fix was simple.
- Added
/llms.txt: We wrote a clear, 3-paragraph summary of Ghost for agents. - Updated Metadata: We injected a rich JSON-LD schema into our
layout.tsx. - Opened the Gates: We updated
robots.txtto explicitly allowGPTBotandCCBot.
The Result?
We re-scanned. Score: 98/100.
(We're still missing a manifest.json for mobile agents, but hey, nobody's perfect).
Lesson Learned
AI visibility isn't a "set and forget" feature. It's a continuous part of your development lifecycle. That's why we built Ghost—so you don't have to be embarrassed by a "32" when the AI agents come knocking.