Why I Built SitemapToLLMs

I've been building developer tools for years - PHP Unserialize, PHP Serialize, PHP Playground, WP Admin Online, and others. They're simple, focused tools that solve one problem well.

But recently, I noticed something shifting in how people discover and use these tools.

More traffic was coming from AI-assisted workflows. People weren’t just Googling “unserialize PHP online” anymore — they were asking ChatGPT, Claude, or Perplexity for recommendations.

And that got me thinking:

How do LLMs actually understand what my sites do?

The answer was uncomfortable.

They don’t — not really. They infer. They scrape fragments. They fill in gaps. Sometimes they hallucinate the rest.

The llms.txt Moment

When I first came across the llms.txt specification, it immediately clicked.

It felt like what robots.txt did for search engines — but for AI.

A structured file at your site root that tells LLMs:

• What the site is about
• Which pages matter
• How everything is organised

The problem was obvious.

Nobody is going to manually maintain an llms.txt file for a site with hundreds of pages. And even if you write one, it goes stale the moment you publish new content.

But I already had sitemaps on all my sites.

Sitemaps already describe structure.

The leap from sitemap → llms.txt felt natural.

Building It

The first version was intentionally simple:

Paste a sitemap URL → get a formatted llms.txt.

Technically correct. Practically useless.

Auth pages leaked through. WordPress archives appeared. Duplicate URLs showed up. The output was just a flat list — no structure, no signal, no prioritisation.

So I iterated.

• URL filtering to remove noise (login pages, admin routes, archives)
• Semantic section naming (turning slugs into readable headings)
• Automatic grouping of legal pages
• Priority-based ordering so important content appears first

With each improvement, the output became more meaningful.

It stopped being a URL dump and started becoming something that genuinely helps an LLM understand a site's purpose and structure.

Making It Sustainable

From building other tools, I’ve learned something simple:

“Free forever” doesn’t pay server bills.

But I also didn’t want to hide the core functionality behind a paywall.

So I landed on something fair:

Free

• Up to 5 sites
• Manual generation anytime
• 100 URLs per site

Pro — $0.99/site/month

• Up to 50,000 URLs
• Automated daily/weekly/monthly regeneration
• Email notifications

The Pro tier solves a real problem.

An llms.txt file is only useful if it stays current.

If you publish weekly content, your AI-facing structure should update automatically — not rely on you remembering to regenerate it.

What I Learned

Building SitemapToLLMs reinforced something I keep rediscovering:

The best tools automate the thing you’d otherwise forget to do.

Nobody wakes up thinking:

“I should update my llms.txt today.”

But everyone wants their site to be discoverable by AI.

The AI discovery landscape today feels a bit like SEO did 15 years ago. Early, undefined, slightly chaotic.

llms.txt might not be the final standard.

But the underlying need — making websites machine-readable for LLMs — isn’t going away.

If anything, it’s becoming foundational.

If you have a website, you’ll likely need some structured AI-facing layer.

And if you don’t want to maintain it by hand, that’s exactly why I built 👉 SitemapToLLMs.