A raw response is only one view
A request may return a small HTML shell whose content appears after JavaScript runs. This explains why a raw-only scan sees little text. It does not prove that Google or another rendering-capable system sees the same result.
Conversely, the presence of text in HTML does not prove it is visible to a user. CSS, overlays and interactive state can change the experience.
Compare meaningful information
A useful rendered comparison should inspect the page title, headings, principal text, links, structured data and interactive controls. Record what changed and which view produced each observation.
Avoid a vague claim that a site is “AI unreadable.” State the finding narrowly: for example, no principal text was available in the raw HTML response, while a browser comparison has not yet been performed.
Review form semantics separately
A button with visible text and a properly associated label are useful structural signals. They do not demonstrate keyboard operation or successful task completion. Test navigation, focus, validation and dialogs in a browser with assistive technology where appropriate.
Our agent readiness checker inspects raw control names. It explicitly reports that interactive behavior and computed accessibility are untested.
Choose a proportionate change
If key information is needlessly delayed, consider delivering that content in the initial response. Verify the behavior against the actual systems that matter to your site. Do not rewrite a working application merely because a raw-fetch heuristic produced a warning.
Start with a website readiness scan, read its limitations, and collect rendered evidence before making a larger change.