Do AI crawlers (GPTBot, ClaudeBot) execute JavaScript?

Most don't. In an SPA the content arrives after client render, so a crawler that doesn't run JS sees only an empty #app in the raw HTML — not the article body. The fix is to bake the content into static HTML at build time.

Do you need SSR to prerender an SPA blog?

No. Full SSR is overkill for a small blog. A little Node script that runs postbuild, takes dist/index.html as a template, and writes dist/blog/ /index.html per post is enough — body, correct head, and JSON-LD in static HTML.

Does prerendering break the JS user experience?

No. Vue takes over #app on mount; the static content only matters while a crawler reads the raw HTML. nginx serves the static file redirect-free with try_files.

← Blog · June 20, 2026

Making an SPA blog visible to AI engines: build-time prerender

This blog is a Vue SPA. The problem: most AI crawlers — GPTBot, ClaudeBot, PerplexityBot — don’t run JavaScript. So when they hit /blog/<slug>, all they see in the raw HTML is an empty <div id="app">. My most citable content — posts full of code and numbers — was invisible to them.

Solving it without SSR

Standing up full SSR (moving to Nuxt and so on) is overkill for a six-post blog. There’s a smaller path: prerender on postbuild. A Node script that runs after the Vite build takes dist/index.html as a template and writes a static page per post:

const template = readFileSync("dist/index.html", "utf8");
let html = template
  .replace(/<title>.*?<\/title>/, `<title>${post.title}</title>`)
  .replace("</head>", `${blogPostingJsonLd}</head>`)
  .replace(/<div id="app">[\s\S]*?<\/div>/, `<div id="app">${articleHtml}</div>`);
writeFileSync(`dist/blog/${post.slug}/index.html`, html);

The result: dist/blog/<slug>/index.html — body, per-route canonical, and BlogPosting/FAQPage JSON-LD all in raw HTML, no JS needed. nginx serves it redirect-free with try_files $uri $uri/index.html; Vue takes over #app on mount, so JS users notice nothing.

Why it works

A crawler that doesn’t run JS now reads real text instead of an empty shell. Google’s JavaScript SEO guide talks about the cost of rendering; most AI engines never pay that cost — the rendered content has to be there for them already. There’s no prerender for an unknown slug, so nginx falls back to index.html and the SPA shows its own 404.

A tiny postbuild script solved the problem a whole SSR migration would have — exactly enough at this blog’s scale.