What AI assistants can and can't see on your website

A website can look perfectly good to a human visitor and still give an AI assistant almost nothing to work with. The gap is not about design quality. It is about whether your facts exist in a form a machine can extract without guessing.

Text inside an image is invisible

This is the single most common problem we see on small business sites in the region. Opening hours in a graphic. A menu as a photographed page. A phone number baked into a banner. A person reads all of it instantly; a crawler reads none of it. If a fact matters, it needs to exist as text on the page — the image can stay, but it cannot be the only copy of the information.

Content that only appears after interaction is unreliable

If your service list only exists after a script runs, or your FAQ answers are injected into the page only once someone clicks, you are betting that every crawler executes JavaScript patiently and clicks things. Some do. Many do not, and the ones that do may not wait. Content that is present in the initial HTML is content you can count on.

This is why the FAQ on this site is built with plain HTML disclosure elements: only one answer is visible at a time, but every answer is present in the page source whether it is open or closed.

Structure carries meaning that prose does not

"We're open until 6" in a paragraph is a sentence. The same fact in structured data is a machine-readable opening hours field attached to a business of a specific type at a specific address. The second one can be quoted with confidence. The first has to be inferred, and inference is where assistants get cautious or get it wrong.

• Use the schema.org type that matches what you actually are, not a generic fallback.
• Mark up the things customers ask about: hours, location, service area, services offered, price range where you publish it.
• Keep a real heading hierarchy. Headings are how a machine understands what a section is about.
• Give the page one clear subject. A page trying to be about five things is quotable on none of them.

You may be blocking the crawlers outright

Some site builders and security plugins block AI crawlers by default, or lump them in with scrapers. If GPTBot, ClaudeBot, PerplexityBot or Google-Extended are disallowed in your robots.txt, no amount of good structure matters — you have opted out of the thing you were trying to win. It costs nothing to check, and it is worth checking before anything else on this list.

What good looks like

Facts as text, in the initial HTML, marked up with the right structured data type, organised under honest headings, on a site that lets the crawlers in. That is most of it. None of it guarantees a citation — that call belongs to the answer engine, and no vendor controls it — but it is the difference between being a candidate and not being in the running.

Crawler names and conventions here reflect research as of mid-2026 and change regularly. We re-check ours quarterly and re-check clients' as part of their Care Plan.