Why AI can't always read your site
Here's a thing most business owners never get told. When ChatGPT or Perplexity go to read your website, they don't see it the way you do. They don't open a browser, wait for your slideshow to fade in, and admire your animations. They grab the raw code that comes back on the very first request, and they read that. Whatever hasn't loaded yet, they usually don't see.
That matters more than it sounds, because a lot of modern websites build most of their content after the page arrives, using JavaScript. Your beautiful homepage might look full to you and be almost empty to an AI. And this isn't a rare edge case – it's how the big AI crawlers work right now.
Read that middle stat again. The AI downloads your scripts, then doesn't run them. So if your prices, your service list or even your business name only appear once JavaScript has done its work, the AI is looking at a page where those things simply aren't there yet. Google's own crawler can run JavaScript, so you can look perfectly fine on Google and still be a blank to ChatGPT. That gap is exactly where good businesses quietly disappear.
The good news: making a site readable to AI isn't some dark art. It's mostly about putting the important stuff in plain, real text and getting out of the AI's way. Let's walk through how.
The nine steps, one by one
You don't need all nine done today, and you don't need to be technical to understand them. Work top to bottom – the earlier steps carry the most weight. If you only ever do the first three, you'll already be ahead of most of your competitors.
That's the whole method. Now let me slow down on the three that people get wrong most often, because getting these right is where the real difference shows up.
Speak the machine's language
Step three, structured data, is the one that sounds scary and isn't. Schema.org is a shared vocabulary that lets you label the things on your page so a machine knows exactly what they are. "This is the business name. This is the phone number. These are the opening hours." You write it once, tuck it into the page, and it never shows to your human visitors.
Structured data = a small block of code that translates your page into a machine-readable format, so AI doesn't have to guess what your words mean.
Why bother, if AI can already read plain text? Because guessing is where it goes wrong. Schema markup "translates human-readable content into a machine-readable format," and by explicitly labelling your content it makes your business easier for AI to interpret and, crucially, easier to recognise as you rather than some similarly-named firm two towns over. Organization schema, in particular, anchors your brand identity so AI systems can consistently reference it.
I'll be honest about the ceiling here, because I hate hype. Structured data is not a magic switch that forces the AI to quote you. It raises the odds and lowers the confusion – nothing more, nothing less. That's still well worth having. Clean, consistent facts written as real text do the heavy lifting; schema is the sturdy label that makes those facts harder to misread.
Schema doesn't say anything new about you – it just labels what's already there so AI can't mix you up with someone else. Get your facts right first, then wrap them in schema.
Write the way people actually ask
Here's the shift that changes everything about how you write. People don't type keywords at an AI. They ask full questions, the way they'd ask a friend. "Who does gutter cleaning near me and how much is it?" So the pages the AI loves to quote are the ones that answer real questions plainly and quickly, in a way it can lift straight into its reply.
That's what "answer-first" means. Lead with the answer, then earn the reader with the detail. Flip your old habit of a long, moody intro before you get to the point. The AI, and honestly your busy human visitor, both want the point first.
Try this: turn a section heading into the exact question a customer would ask, then make the very next sentence a complete answer that would still make sense pulled out on its own.
A page with a clean heading order, plain answers and descriptive text is a page an AI can parse reliably – it can find the section that matches a specific question and lift that part out without getting confused by the noise around it. Marketing fluff is exactly that noise. A wall of "we're passionate about excellence" tells the AI nothing it can use.
Write for the person asking the question, in the words they'd use to ask it. Do that, and you're writing for the AI at the same time.
Danielle Evaland, Founder of EvaraNone of this means dumbing down. It means clarity first. Say the true, specific, useful thing – then add the warmth and the story. If you want the deeper reasons AI sometimes still gets your details wrong even when you've done all this, we unpack it in why AI gets your business facts wrong.
Open the door: robots.txt and llms.txt
You can do everything above and still be invisible for a dull reason: you accidentally shut the AI out, or you never rolled out the welcome mat. Two small files handle both.
| robots.txt | llms.txt | |
|---|---|---|
| What it does | Tells crawlers what they're allowed to read | Hands AI a clean summary of your best content |
| Where it lives | yoursite.com/robots.txt | yoursite.com/llms.txt |
| Format | Simple rules | Plain markdown text |
| Get it wrong and… | You block the AI without meaning to | The AI has to wrestle your messy HTML alone |
The robots.txt is the bouncer. Make sure it isn't turning away the AI crawlers you actually want reading you. A single stray line can quietly block GPTBot or ClaudeBot from your whole site, and you'd never see a warning.
The llms.txt is newer and I genuinely like it. Proposed in 2024 by Jeremy Howard, it's a simple markdown file that gives AI tools a tidy summary of your site with links to your key pages. It exists for a very human reason: AI has a limited "attention span" (a small context window), and turning a cluttered page full of navigation, ads and scripts into clean text is "both difficult and imprecise." An llms.txt skips all that and hands over the good bits already flattened into plain text. It's a proposal, not a law yet, but big names like Anthropic and Cloudflare already publish one – and it costs you almost nothing to add. We go deeper in what is an llms.txt file.
One caution: llms.txt is voluntary, so not every AI reads it yet. Treat it as a helpful bonus on top of clean HTML and good facts, never as a substitute for them.
Your quick AI-readable checklist
Run your homepage past these. Every "no" is a place an AI might be losing the plot about your business – and every one is fixable.
- Can you copy your key facts as text? Highlight your name, services and phone with your cursor. If you can't select and copy them, they may be trapped in an image or a script.
- Does the page have a clear heading order? One main heading that names what you do, then tidy sub-headings underneath in a logical flow.
- Is there structured data? A LocalBusiness or Organization schema block labelling your core details for machines.
- Do your headings answer real questions? With a plain, self-contained answer in the first sentence or two beneath each one.
- Do your facts match everywhere else? Same name, address, phone and hours on Google, Facebook and directories.
- Are the crawlers allowed in? Your robots.txt welcomes GPTBot, ClaudeBot and PerplexityBot, and you've added an llms.txt.
- Do images and links describe themselves? Plain alt text on pictures, and link text that says where it leads.
If a few of those are a "no", that's normal – and it's genuinely good news, because it means the reason you're being skipped is fixable, not some mystery. My honest take after doing this for a lot of businesses: readability is the cheapest, fastest win in the whole AI-visibility game. You're not fighting anyone. You're just clearing the fog so the AI can see what was always true about you.
If you'd rather have it built in from the ground up instead of retrofitted, that's exactly what our websites built for AI do – readable to the machines, and lovely for the humans who buy from you.

By Danielle Evaland



