Technical guide ·
Does your website’s platform affect AI readability?
Your technology stack tells you where to look. The page response tells you what is actually there. And even perfectly readable content can still be missing the answers customers are looking for.
Download this article (.md)Vercel has the tools to make content readable
A site hosted on Vercel can return complete, useful HTML. It can also return a thin page that relies on JavaScript and later API requests for important content. Hosting alone does not tell you which one you have.
Next.js supports server rendering and prerendering. Its documentation explains that the initial HTML can include both Server and Client Components. Using “use client” is therefore not evidence that a page is invisible without JavaScript. Content fetched only after the browser loads still needs checking. Next.js rendering documentation
The relevant question is not “Does this use Vercel?” It is “Which important facts are present in the response this page returns?”
JavaScript can change what a crawler receives
In research published on December 17, 2024, Vercel and MERJ reported that the tested OpenAI, Anthropic, Meta and Perplexity crawlers did not execute JavaScript. Googlebot and Applebot were rendering-capable counterexamples. This was an observation of those systems during that study, not a permanent specification for every AI product or retrieval path. Vercel and MERJ research
Fetching a JavaScript file is not the same as executing it. Equally, an AI answer mentioning your business is not proof that a particular crawler rendered your homepage. The answer may rely on another page, an index or other sources.
What to check in common stacks
Use these checks to find where important information enters the page.
| Stack or pattern | Useful check |
|---|---|
| Vercel / Next.js | Compare initial HTML with the rendered page. Inspect client-fetched content and pages where prerendering has been disabled. Do not classify Client Components as missing content automatically. |
| React, Vue or Angular applications | If a page is client-rendered, check whether useful text exists before scripts run. The framework name alone does not establish the rendering mode. |
| Nuxt | Universal rendering is the default, with client-only and hybrid modes available. Check the actual route configuration and response. |
| Shopify themes and apps | Check whether product information, variants, qualifications and app-provided answers are available in the initial page, rather than only in a widget. |
| Adobe Experience Manager / headless CMS | Inspect the frontend delivery architecture and individual components. A CMS storing a fact does not establish when the frontend publishes it. |
| WordPress, Webflow and enterprise commerce | Inspect real product, article, category and policy pages. Navigation volume, templates and widgets are reasons to review output, not proof of poor AI visibility. |
Angular supports client, server and prerendered routes. Nuxt defaults to universal rendering. Adobe documents JavaScript-driven SPA implementations and their rendering considerations. These options reinforce why implementation matters more than the logo on the stack. Angular, Nuxt, Adobe
Shopify recommends rendering essential content in Liquid and HTML instead of relying on client-side JavaScript. Start with product details and answers supplied by apps: check whether they arrive with the page or require a later browser request. Shopify guidance
Test pages, not just homepages
- Choose a homepage, product or service page, category page, article and policy page where applicable. Record the URL, time and extraction settings.
- List the important facts first: names, capabilities, prices, units, eligibility conditions, supporting links and qualifications.
- Inspect the initial HTML, then compare a browser-rendered version. Record which facts require JavaScript and which remain missing.
- Inspect the cleaned result. Check for lost facts, detached footnotes, changed table relationships or missing category links. A shorter result is not automatically more useful.
- If a request fails, report the observed status or challenge. An anonymous fetch cannot establish that OpenAI, Google or another specific service is also blocked.
The public Website Preview can help inspect a URL with or without bounded browser rendering. It is not a named AI crawler’s session, a complete site audit or an AI visibility score. Resource restrictions and time limits can leave content incomplete.
Readable doesn’t mean complete
Making existing website content accessible is the first problem. A page can be server rendered, easy to crawl and cleanly structured—and still lack the answer a customer is looking for.
Product details may be spread across support documentation. Implementation answers may live with sales teams. Important differences between products may be obvious internally but absent from the website.
Consider a product page that loads perfectly. Does it explain who the product is for, what is included, which systems it works with, its limitations and how to get started? Rendering and format conversion cannot supply facts the business has never published.
- Can machines retrieve the information you have already published?
- Have you published the information you want customers and AI systems to find?
Server rendering, prerendering and browser rendering primarily address the first question. The second is a content and publishing problem: gather the answers, confirm them with the people responsible and publish them with supporting links.
Choose the fix that matches the finding
If essential facts are absent from initial HTML, server rendering, prerendering or moving the content into the server-delivered page are valid options. If the information itself is missing or inaccurate, update the source content. If the problem is access, investigate the actual security rule instead of removing protections broadly.
Managed rendering is also available from infrastructure providers. Cloudflare’s crawl API can discover pages and return rendered HTML, Markdown and structured JSON. Rendering and conversion are not exclusive Optiview capabilities. Cloudflare Browser Rendering
Optiview addresses both layers. It turns supported website content into a clearer machine-readable source and gives authorized teams a way to add approved brand details and answers that the existing page does not explain clearly. Your team reviews the combined result before publishing. The web team establishes delivery; content teams can keep that source current without changing application code for every update.
Readable content still needs a delivery path
A custom API integration lets an authorized server request the published result and control delivery. Neither method guarantees indexing, citations or better answers. For Google Search and its generative features, Google says special AI files, markup and Markdown are not required. Google Search guidance
Start with the page, then close the gaps
Your technology stack can tell you where to look. The response tells you what is actually there.
Next.js, React, Angular and Shopify can all support accessible pages. Important information can also arrive only after browser execution or remain scattered across applications, documents and teams.
Inspect representative pages, identify what machines can retrieve, and determine whether the important answers are actually there. Then fix the specific gap: how the content is delivered, what the brand has published, or both.
The goal isn't to optimize for a particular crawler. It's to make sure your owned source says what it needs to say—and that machines can actually get to it.