How to tell whether a crawler can see the text on a JavaScript website

Two representations of the same page

The initial HTML is what the server sends when an address is requested. The rendered page may be supplemented with API data, widgets and scripts. Google describes separate JavaScript retrieval and rendering stages; resource availability and errors affect the outcome. Other clients do not have to replicate Google’s capabilities.

Compare meaning, not file size. A large HTML response may consist of navigation, while a short one may already contain a complete answer. Text inside a JavaScript variable does not necessarily mean that a standard extractor will recognize it as the main content. A canvas or an image containing text also needs a separate assessment.

What to check without rebuilding the entire project

Choose a service page, an article and an example product page. Open the initial response and look for the heading, description, price or terms, key facts and links. Then compare it with the browser. Record which important elements appear only after loading.

If the useful text itself is missing, discuss server rendering, prerendering or another supported delivery method with a developer. Do not enable a separate version with exaggerated promises for bots. Visitors and crawlers should receive a consistent meaning. For application routes, also test opening the URL directly, not just navigating from the homepage.

The AURA report and criteria for a completed fix

The AURA audit checks the retrieved HTML and extracted information. This is a limited technical check, not a browser simulation of every AI service. A low detected word count is therefore a reason to open the response and investigate, not a final verdict on the website’s technology.

Give the agent a specific page and a list of missing facts. It needs source code or CMS access to make the edit. After deployment, repeat the request and confirm that the main text is present and the interface still works.

The content should be readable through a direct request, describe the same product and contain the required links. Save an example response before and after the change. Then observe discovery and citations separately. Changing the rendering method removes a technical dependency; it does not make any text relevant to every question or establish a brand’s position in an answer.

For a published review, see the Pupupul audit. It shows checks of accessible content and specific obstacles. It is an example of technical work, not confirmation of an AI recommendation.