How to check if AI can read your site: a 5-minute audit

When someone asks an assistant about your topic, the model doesn’t open your site the way a person does. It fetches your HTML, skims for the parts it can parse, and decides — in milliseconds — whether your page is worth reading, summarizing and citing. Most pages pass some of that gauntlet and quietly fail the rest, leaving citations on the table without ever knowing.

Here’s the good news: the checks that matter are few, and you can run all of them in about five minutes. Work through these four in order.

1. Can AI crawlers actually reach you?

Before anything else, the engines have to be allowed in. A single line in robots.txt can shut out the exact bots you want — GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, and Google’s Google-Extended training token. Open yoursite.com/robots.txt and confirm none of those agents (or *) are hit with a blanket Disallow: /. It’s the most common own-goal we see, and the easiest to fix.

2. Is your content readable without JavaScript?

Many crawlers don’t execute JavaScript. If your article text only appears after a client-side render, the model may see an empty shell — and guess. View your page with JS disabled (or just “View Source”): the words that matter should be right there in the HTML. Server-render the important content; don’t make the answer engine wait for a framework that never runs.

3. Do you tell AI how to summarize and cite you?

Models read your markup, so remove the ambiguity. The AI Graph tags declare, in plain terms, what your page is and how you want to be attributed:

<meta property="ai:type" content="article">
<meta property="ai:title" content="Pour-over coffee, done right">
<meta property="ai:summary" content="A step-by-step pour-over method…">
<meta property="ai:cite_as" content="Craft & Quill">

These aren’t a trick to rank higher. They’re a way to be represented accurately instead of paraphrased into something you didn’t say.

4. Do you have structured data and a canonical URL?

A JSON-LD block (Article, Product, Recipe) gives the model typed facts it can trust, and a <link rel="canonical"> tells it which URL is the real one so citations don’t scatter across duplicates. Both are small additions with outsized clarity.

Run it automatically

You don’t have to eyeball all of this. The free AI Discoverability Score scans any URL exactly the way an AI crawler would — robots access, readable text, structured data, AI Graph tags and canonical — and gives you a 0–100 score with the specific fixes that move it. No signup, and it never stores your page. Start with one URL that already answers a real question well, close the gaps, and watch how the assistants treat it.