Guides and findings · Guide

Can AI read your site?

Six checks, a minute each. If an assistant can’t read you, it can’t name you, and it certainly can’t pick you.

6checksA minuteeachFreeno tools needed

01

Let the crawlers in

If your robots.txt blocks AI crawlers, assistants answer from what others say about you instead.

Check it in a minuteyourbrand.com/robots.txt
Good looks like

Look for lines naming GPTBot, ClaudeBot or PerplexityBot followed by “Disallow: /”. No such line, or “Allow”, is good.

The common miss

A blanket block added years ago, for a different reason.

02

Publish a sitemap

A sitemap tells any crawler which pages exist and which are new.

Check it in a minuteyourbrand.com/sitemap.xml
Good looks like

A list of your real pages, with recent dates. Better still, named in robots.txt.

The common miss

A sitemap full of old or redirected pages.

03

Add an llms.txt file

A short plain-text guide to your site, written for language models. It’s a proposed convention, and not every assistant reads it yet, but it costs an hour.

Check it in a minuteyourbrand.com/llms.txt
Good looks like

Who you are in one paragraph, then links to the pages that matter, each with a line on what it answers.

The common miss

A copy of the homepage, with no links.

04

Mark up the facts

Structured data states your name, products, prices and questions in a form machines read without guessing.

Check it in a minuteAny page → view source → search “application/ld+json”
Good looks like

Organization on every page, Product on product pages, FAQPage where you answer questions.

The common miss

Markup that says something the page doesn’t.

05

Put a source by every claim

Assistants are wary of claims only you make. A claim with a source is one they can repeat.

Check it in a minuteYour three biggest claims
Good looks like

Each one links to its proof: a published figure, a named source, an independent review.

The common miss

“Industry-leading”, with nothing behind it.

06

Say what it costs

Buyers ask what it costs early. If your site won’t say, the answer comes from somewhere else, or nowhere.

Check it in a minuteAsk an assistant what you charge
Good looks like

A price, or a price range, on a page an assistant can read, with the date it was set.

The common miss

“Contact us for pricing”, and nothing else.

Or check all six on every publish.

The Website Engine runs these checks before a page goes live: dead links, imagery, AI agents and figures. Then a receipt for exactly what went live.

Checked for AI agentsA receipt for what went live

Six out of six is the start.

Passing all six means an assistant can read you. It doesn’t mean it picks you: being read is the Found stage, one of seven. The free score shows the other six.