People no longer search for a business only on Google; they ask ChatGPT, Perplexity, Claude or Gemini. These tools build their answers from the pages they can read. If your site cannot be read, it will not be named in the answer, however good it is. Here are the changes we made on our own site (vibetobrands.com) for that reason; you can run the same checks on yours.
1. The content has to be in the page source
Most AI search crawlers do not run the JavaScript on a page; they read the raw HTML the server sends. A site that fills in its text with JavaScript after loading looks empty or half-finished to them. Our own site started out like that: the English pages were translated in the browser with JavaScript, so a crawler reading the raw HTML saw Turkish text at the English address.
The fix was to write each page's text, prices and FAQs straight into the HTML. A build step runs before publishing and produces the finished page for each language. Visitors see no difference; crawlers read the whole page.
The easiest test for your own site: right-click the page, choose “View page source” and search for an important sentence. If it is not in the source, AI search most likely cannot see it either.
2. Every language gets its own address
If a bilingual site switches language with a button but keeps the same address, crawlers only ever see one language. Our Turkish pages live at the root and the English ones under /en/, with English addresses. Each page points to its counterpart in the other language with an hreflang tag.
3. Structured data: a summary for machines
We also give the page's information as a schema.org block that machines can read directly: the service name, description, price and currency, and where the page sits in the site. On the Turkish page the price is in lira, on the English page in dollars, because structured data has to say the same thing as the visible page. We use FAQ markup only where the questions are actually visible on the page.
4. llms.txt and robots.txt
llms.txt is a short summary of a site written for AI assistants: what you sell, your prices, your key pages. It is a proposal rather than a standard so far, but it is easy to write and it describes your site correctly to the tools that read it. Ours is at vibetobrands.com/llms.txt.
robots.txt tells each crawler where it may go. When writing separate rules for AI crawlers, watch for three traps: once you give a crawler its own block, the general (*) rules no longer apply to it; a blank line inside a block splits it in two; and the order of the allow line changes the outcome for some crawlers. A badly written robots.txt can shut your site to tools you never meant to block.
5. The price has to match everywhere
When we change a price in the panel, the price on the page, the price in the structured data, llms.txt and the site assistant's knowledge are all updated together. AI tools read the same fact from several places; if one of them is stale, they may repeat the stale one.
What this does not guarantee
These steps make your site readable; they do not guarantee a position in search results or in AI answers. Visibility depends on how well your content answers the question, and on your competitors. But for a site that cannot be read, that race never starts.
In our website packages the full setup comes with the Business and Comprehensive packages, and the basic version with the Starter package. The sites we build are published the same way as this one, as raw HTML.
