More than one in three popular JavaScript framework sites I tested lost important content without JavaScript when I compared the HTML their servers send with the rendered page. When I re-ran the same homepages in November 2026, only three of the 83 sites I could compare both times got a different verdict.
Google copes with these pages, but most AI crawlers read only what your server sends straight away. Below I'll show you how to check JavaScript SEO on your own site in a few minutes and what to fix first. You'll find the scripts and data in the research pack.
What is JavaScript SEO?
JavaScript SEO is making sure search engines and AI crawlers can fetch, read and index the content a site builds with JavaScript. Every page you've got exists in two versions. The server HTML is what curl or "View source" (Ctrl+U) shows you, and the rendered DOM is what the Elements panel shows once the scripts have run.
How far apart the two are depends on how your framework renders the page.
| Rendering mode | Where the HTML is built | What a crawler without JavaScript gets |
|---|---|---|
| Client-side rendering (CSR) | In the visitor's browser | An empty shell, often just <div id="root"> |
| Server-side rendering (SSR) | On the server, for each request | The full page, which JavaScript then brings to life (hydration) |
| Static site generation (SSG) | On the server, once at build time | The full page from a file |
| Hybrid | Some parts on the server, some in the browser | Whatever the server part contains |
How Google renders JavaScript pages
Google handles JavaScript pages in three phases, with a rendering queue sitting between crawling and indexing. According to Google's JavaScript SEO basics, a page can wait in that queue for a few seconds, "but it can take longer than that". Then a headless Chromium runs the scripts.
I'd pay more attention to two other details there. On the first pass Google finds links only in <a> elements with an href attribute, so navigation injected by a script waits for the second pass. Google also calls server-side rendering or pre-rendering "still a great idea", because it's faster and because not every bot can run JavaScript.
So Google usually does see your client-rendered content, only later and after more steps where something can break, such as a timeout or a script error. And Google is just one of the crawlers that visit your site.
AI crawlers read only the server HTML
Most AI crawlers read the HTML your server sends and skip the JavaScript. Vercel and MERJ looked at AI crawler traffic across Vercel's network in The rise of the AI crawler. They found that none of the major AI crawlers render JavaScript, apart from AppleBot and Gemini, which uses Googlebot's infrastructure. The crawlers behind ChatGPT and Claude do download script files, but they never execute them.
That analysis is getting old, so I looked for a newer one. Resoneo set up a test page with 15 ways of injecting content and asked assistants to read it.
In Resoneo's test, ChatGPT and Claude read only the static HTML when a user asked them to open the page. Gemini's live fetch did the same, and Perplexity's fetch came back empty. Bing Copilot and Grok did run JavaScript, though Copilot waited only a fraction of a second.
Translation? If your product descriptions or article text appear only after JavaScript runs, the assistants your customers ask don't have anything to quote.
One in three JavaScript framework sites loses content without JavaScript
More than one in three framework sites in my sample lost important content in the server HTML, in both runs. That means a key tag such as the H1 or canonical showed up only after JavaScript, or the server HTML held less than half of the text or links.
Text went missing most often, then internal links and the H1. The title survived almost everywhere, and I think that's why the problem hides so well: your browser tab and search snippet look fine while a bot gets an empty page.
The repeat changed one thing in the method, because the first run fetched pages as desktop Chrome and the second as my research crawler. Only three sites flipped, so the result holds for this sample.
React SEO: client-side rendering loses more than server rendering
React SEO problems come from client-side rendering, and Vue and Angular have the same problem. In my sample more than half of the sites on plain React, Vue or Angular lost important content without JavaScript. On Next.js, Nuxt, SvelteKit and Gatsby, which render on the server by default, it was fewer than three in ten.
You can still get it wrong on a meta-framework. Next.js on its own did no better than the sample as a whole, with more than a third of its homepages losing important content. A few were empty shells because they loaded their content in the browser.
The server-rendering group looks better mainly thanks to Nuxt. The framework hands you server rendering, but each route still has to use it.
Five common causes of JavaScript SEO problems
A few patterns explain most of the gaps I saw, and you can spot each one in your server HTML.
- Data fetched after the page loads. A component that calls an API in
useEffect(React) oronMounted(Vue) ships an empty container from the server and fills it in the browser. Look for a heading with no paragraph under it, or an empty<ul>. - Metadata set by a browser-only head library. In a plain SPA the title and canonical get written by a script, so the server HTML carries the default ones from
index.html. If several different routes return the same title incurl, this is your cause. - Navigation without real links. A menu built from
<div>or<button>elements with click handlers looks identical to a user and gives a crawler nothing to follow, because a bot doesn't click, it follows the URL inhref. - Content behind "load more" or infinite scroll. Products and older posts that arrive only on scroll or on click never reach the server HTML, so give each page of results its own URL, such as
?page=2. - A firewall that blocks AI crawlers. It isn't JavaScript, but you get the same blank page. Vercel and Cloudflare both ship ready-made firewall settings that block known AI bots, so ask whoever runs yours what it's set to.
For links, the change in code is small, and for a crawler it's everything:
<!-- A crawler can't follow this -->
<div class="nav-item" onclick="router.push('/pricing')">Pricing</div>
<!-- A crawler can follow this, and the router still handles the click -->
<a href="/pricing">Pricing</a>
How to check JavaScript SEO on your own site
The quickest JavaScript SEO check takes a few minutes and no paid tools, and I run these steps in order.
- Fetch the server HTML. Run
curl -sL -A "Mozilla/5.0" https://yoursite.com/ > page.html, count the H1 tags withgrep -o "<h1" page.html | wc -l, then search the file for a sentence from your main content. If the sentence isn't there, a crawler without JavaScript misses it too, and if you got a bot-check page instead, your firewall is answering for the site. - Compare "View source" with "Inspect". View source shows the server HTML and the Elements panel shows the rendered DOM. Anything that appears only in the second one depends on JavaScript.
- Turn JavaScript off. In Chrome DevTools press Ctrl+Shift+P (Cmd+Shift+P on a Mac), type "Disable JavaScript" and reload. What's left is roughly what an AI crawler reads.
- Check what AI crawlers may fetch. Our free AI visibility checker fetches one page the way an AI crawler does, skipping JavaScript, shows how many words of main text the server HTML holds, and checks your robots.txt for 14 crawlers.
- Check what Google rendered. In Search Console go to URL Inspection > Test live URL > View tested page > HTML. That's the version Google rendered.
Run step 1 on one URL from each template, because the homepage is often static while the trouble starts on product pages or posts. This loop prints how many links each template sends before JavaScript:
for u in / /pricing /blog/your-best-post /products/your-best-product; do
printf "%s " "$u"; curl -sL -A "Mozilla/5.0" "https://yoursite.com$u" | grep -o "<a [^>]*href" | wc -l
done
I ran the same comparison on cometweb.io, which is built on SvelteKit. The homepage's server HTML has every key tag and every internal link. About a fifth of the text appears only after JavaScript, and it all comes from the interactive Insight demo and the cookie banner.
SEO for single-page applications: how to fix it
SEO for single-page applications gets fixed on the server, because more patches in browser code only move the problem around. Every page you want indexed or cited should leave the server with its content and links already in the HTML, and here's the order I'd work in.
- Render indexable routes on the server. Use SSR or static generation for every page you want to rank, from products to articles. Next.js and Nuxt do this by default, and Angular has its own SSR package. A plain React or Vue SPA usually needs a move to Next.js or Nuxt, or at least prerendering its public pages at build time.
- Put metadata in the server template. The title, description and canonical belong in the first HTML response. Google can read a canonical set by JavaScript, but its docs ask you not to let JavaScript change the one in the original HTML. If you're writing the head block by hand, the meta tag generator gives you a clean starting point.
- Use real links. Each navigation and pagination item should be an
<a>element with a real path inhref, such as/products, and History API routing. Click handlers and#/productsfragments give crawlers nothing to follow, which is why my internal link audit starts from the links a crawler can see in the source. - Return honest status codes. A missing product in a SPA often returns 200 and the app shell, and the "not found" text appears only once a script runs. Google treats that as a soft 404, so ideally the server returns 404 or 410 straight away. In a pure SPA, Google suggests a JavaScript redirect to a URL that returns 404, or a
noindextag added by JavaScript.
SSR, SSG, ISR or CSR: which rendering strategy to pick
Most JavaScript SEO problems come down to the wrong rendering mode for a given page type. One app can mix several methods, so you'll want to decide per template.
| Strategy | What a crawler gets in the first HTML | When it fits | What to watch |
|---|---|---|---|
| SSG (static generation) | All the content | Blog, docs, landing pages | You rebuild the site after a content change |
| ISR (static with revalidation) | All the content | Product catalogues, large content sites | Content can be stale for a short while |
| SSR (server rendering) | All the content | Pages that depend on live data | Server cost and time to first byte (TTFB) |
| CSR (browser only) | An empty shell | Logged-in dashboards, internal tools | Not for pages you want in Google or AI answers |
Serving a prerendered copy only to bots was the old workaround. Google now calls dynamic rendering "a workaround and not a recommended solution", so I wouldn't add it to a new project, and on an old one I'd treat it as a bridge to SSR or SSG. My starting rule is simple. Anything that should show up in search results or AI answers goes into the server HTML, and JavaScript only improves that page.
How to explain JavaScript SEO to a client or developer
The simplest way to explain JavaScript SEO is to show both versions of the same page side by side. Put the curl output next to the page in a browser and highlight the H1 in each. When it's missing on the left, nobody needs a lecture on rendering queues.
With a developer I skip "SEO is broken" and ask three questions:
- Which routes render on the server, and which only in the browser?
- Where do the title and canonical get set, in the server template or in code that runs in the browser?
- Are the menu and product cards made of
<a href>links, or of click handlers?
The objection you'll hear most is "Google renders JavaScript anyway". It does, but later and only for Google, while the assistants your clients use read the server HTML.
Check your JavaScript SEO fix with an audit
Rendering is one item in a wider technical SEO check, so once your important templates return full server HTML, look at the rest too. CometWeb Insight is our website audit app. One audit covers performance (Core Web Vitals), technical SEO, accessibility (WCAG), security and an estimated carbon footprint, and you get the result as tasks sorted by priority.
After you ship SSR you re-run the audit and compare the results before and after, which matters here because a move to server rendering usually changes your load times too. You can start a free account and run your first audit today.
Final thoughts on JavaScript SEO
The diagnosis takes one curl and one search for your H1. The fix takes longer, because moving a client-rendered app to SSR touches your routing and data fetching.
You don't have to rework everything at once. Start with the curl check on your homepage, best-selling product and most-read article. If one comes back without its main text, put server rendering for that template at the top of the next sprint.
Frequently asked questions
Is Next.js SEO-friendly by default?
Yes, Next.js renders pages on the server by default, and components marked 'use client' are rendered to HTML on the first load too. The trouble starts when you fetch data only in the browser, for example in useEffect, because then the server sends just a loading skeleton. So check each important template with curl even on Next.js.
Do I need server-side rendering for an app behind a login?
An app behind a login doesn't need server-side rendering for SEO, because search engines and AI crawlers can't reach it anyway. Render the public parts on the server, from the homepage and docs to the login and sign-up pages.
Does blocking GPTBot hide my site from ChatGPT search?
OpenAI runs separate crawlers for training and for search, so a robots.txt rule for GPTBot alone leaves OAI-SearchBot free to fetch your pages, and your site can still show up in ChatGPT results. Check which user agents your robots.txt and firewall actually block before you decide what to allow.
Does Google see content loaded by JavaScript?
Usually, but it's delayed, after the page goes through the rendering queue. Google finds script-added links later, and a script error or a blocked file can keep part of the content out of the index altogether.
Is dynamic rendering a good solution?
Google calls dynamic rendering a workaround. It'll work, but you're maintaining two versions of each page and risk their content drifting apart, so if you can, move the pages you want in Google to SSR or SSG.
Sources
- Google Search Central: Understand the JavaScript SEO basics. Crawling, the rendering queue, links, canonical and soft 404s in single-page apps.
- Google Search Central: Make your links crawlable. What Google needs from
<a>elements and thehrefattribute. - Google Search Central: Dynamic rendering as a workaround. Why Google no longer recommends it.
- Google Search Central: Fix Search-related JavaScript problems. Debugging rendered pages and avoiding soft 404s.
- Vercel: The rise of the AI crawler. Which AI crawlers fetch and render JavaScript, from Vercel and MERJ traffic data.
- Resoneo: AI Search crawler test results. What assistants read when a user asks them to open a page.
- OpenAI: Overview of OpenAI crawlers. GPTBot, OAI-SearchBot and ChatGPT-User and what each one is for.
- web.dev: Rendering on the web. SSR, static generation, CSR and hydration compared.
- CometWeb: server HTML vs rendered page on JavaScript framework sites. Both runs, scripts and data.
Disclosure: CometWeb Insight and the free tools at /tools are products of CometWeb, the publisher of this blog.