How to check whether AI can read your website
Five checks you can run yourself to determine whether ChatGPT, Perplexity and other generative engines can actually read your site. None require paid tools.
Before a generative engine can cite your company, it has to be able to read your content. The five checks below can be run without paid tools and take roughly fifteen minutes.
One: with JavaScript off, is the content still there?
This is the critical one. Most AI crawlers do not execute JavaScript, so anything rendered client-side by a front-end framework effectively does not exist for them.
The most direct check is to run:
curl -s https://yoursite.com | grep "a sentence from your homepage"If nothing comes back, that sentence is not in the raw HTML — it is generated after the browser runs JavaScript. AI crawlers never see it.
You can do the same thing in a browser with Ctrl+U (Cmd+Option+U on Mac) to view source, then search for the same sentence.
If this check fails: the fix is server-side rendering or static generation, so the content exists in the HTML before it reaches the browser. This is usually a framework configuration change rather than a rewrite.
Two: does robots.txt lock AI crawlers out?
Open https://yoursite.com/robots.txt and look for these names, and whether they are allowed or disallowed:
GPTBot,OAI-SearchBot,ChatGPT-User(OpenAI)ClaudeBot,anthropic-ai(Anthropic)PerplexityBot(Perplexity)Google-Extended(Google’s AI products)
Plenty of sites added blocking rules years ago out of concern about training data and never revisited them. The result is a site search engines can read and AI cannot.
Worth noting: Google-Extended controls access for Google’s AI products only and has no bearing on search ranking. Blocking it neither helps nor harms SEO, but it does keep you out of AI Overviews.
Three: is there structured data?
Open Google’s Rich Results Test and enter your URL. It lists the structured data types detected on the page.
A company should have at least Organization. Anywhere with a physical address should add LocalBusiness. Service pages should carry Service. Pages with common questions should carry FAQPage.
FAQPage is particularly worth doing, because the question-and-answer format matches how generative engines work — a question paired with an answer is exactly the structure they are looking for.
Four: is your company information identical everywhere?
Generative engines use cross-source consistency to confirm a company genuinely exists. Check that the name, address and phone number match word for word across:
- Your site footer and contact page
- Your structured data
- Google Business Profile
- Company registry records
- Social platform profiles
Common inconsistencies: an address written as “2nd Floor” in one place and “2/F” in another, a phone number sometimes with the country code and sometimes without, a company name sometimes with “Limited” and sometimes without. These read as trivial to a human and erode machine confidence.
Five: just ask
The simplest check comes last. Open ChatGPT or Perplexity and ask the question a prospective customer would actually ask.
Do not ask “what is JXT” — that only tests whether it recognises your name. Ask “which companies in Hong Kong build POS systems”. That is the real customer behaviour.
Record which companies appear and which sources get cited. That is your baseline. Ask again a few weeks after making changes and compare.
After the checks
If all five pass, the technical foundation is sound and the work shifts to the content itself: whether headings are phrased as questions, whether answers come first, whether the specifics can be verified.
If several fail, start with the first two. Until the content can be read at all, no other optimisation matters.