What to do when AI cannot read your site
Your site is fine. It opens, it looks right, people find it. And still, the assistant answering a question about you does not draw on it, or describes you from somewhere else entirely. The reason is usually mechanical: what a machine receives from your address is not what you see on your screen.
How a page arrives empty
None of these is a mistake anyone made. They are ordinary ways of building and protecting a site, and each of them happens to leave a reader that is a machine with nothing to take.
- The page is assembled by scripts. The address returns a near-empty shell, and a browser runs the code that fills it. Something that fetches without running anything receives the shell.
- A bot challenge answers first. Protection that keeps automated traffic out answers a waiting page to anything that does not look like a browser. It happens to well-run sites, and to every agent alike.
- A consent wall stands in front. The real page is behind a click, and there is nobody on the other side to make it.
- The answer is inside an image or a PDF. A price list photographed, a brochure scanned: readable to you, not text to anything else.
If you have no site at all, and want none, that is a different question with its own answer: be present in AI without running a website.
How to check, in a minute
Ask your own assistant to open your address and tell you, in its words, what it found there. If it comes back with your menu and little else, or with nothing at all, you have your answer.
The blunter version, if you are comfortable at a terminal, asks for the page the way a machine does and prints the beginning of it:
curl -sL https://yoursite.example | head -40
If your own words are not in what comes back, they are not in what a machine reads either.
An address that answers in text, always
A Markleex page is served as HTML and, at the same address with a
.md suffix, as the exact text you wrote. Nothing to execute,
nothing standing in front, no wall to click through. That is the contract of
the page rather than a setting you switch on, which is why it holds whatever
else changes.
It works alongside your site: your site keeps doing what it does for the people who visit it, and this is the address that answers when yours cannot.
What to publish there
- What you do, in plain words, and who you do it for.
- The facts locked in an image or a PDF: prices, ranges, opening times, coverage.
- What a visitor asks you before buying, one page per question.
- Where you operate, and since when.
- Your relationships, each published once the other entity confirms it.
How to do it
- Look at what a machine receives from your address, rather than at what your browser shows you.
- Claim your Markleex Name.
- Publish, as plain text, the facts your site cannot deliver.
- Declare your website, so the platform reads it and dates what it saw.
- Add one page per question you are actually asked.
Your own assistant can do the writing: connect it to Markleex and point it at the pages you want carried over.
What it costs
A space is $0.83 a month, with 10 pages included, which covers the facts a site most often keeps to itself. See the pricing.