The Site Loaded Perfectly. View Source Was Almost Empty.

    technical seo
    website audit
    checklist
    structured data
    A laptop on a wooden desk showing a nearly empty browser window, with a mug and a small plant beside it

    The first time you open a vibe-coded site, it looks finished. The hero loads, the buttons work, the domain is live. Then you type the brand into Google and nothing useful comes back, or one result appears with a default Vite title and a description that belongs to a different page.

    That pattern is common on sites built in Lovable, Bolt, Cursor, Claude Code and similar tools. They often ship as single-page apps. The browser fills in the page after JavaScript runs, so a person sees a finished product. A crawler that reads the first HTML response can see a nearly empty document: one generic title, no per-route meta, no sitemap, and links that only exist once the app hydrates.

    Short videos have been making the same point about vibe-coded sites that never appear in Google, often with a long SEO checklist. The reminder is fair, but these sites usually fail earlier than missing alt text. If the HTML shell has no unique titles, no crawlable routes and no sitemap, the later items will not get you indexed. Start with the prompt that audits and fixes SEO on a vibe-coded site, then work through the output in your repo.

    What usually breaks on a vibe-coded app

    The problems cluster around how the app is rendered.

    • Client-side rendering with no prerender. Titles, body copy and links appear only after JavaScript. View source, or curl the URL, and you get an almost empty shell.
    • One title for every route. Home, pricing and blog share the same document title and meta description because the head was set once in index.html and never updated per route.
    • Missing robots.txt and sitemap.xml. Or a sitemap that lists hash routes, prerender URLs, or pages that return an error.
    • No canonical tags, or several host variants (www, bare domain, trailing slashes, index.html) all returning 200.
    • No JSON-LD in the served HTML. Tags injected only in the browser never show up in the first response. Invented star ratings or prices in schema will also cause trouble.
    • Open Graph tags missing or identical on every page, so shares look blank or repeat the homepage.
    • Unknown paths return 200 with a client-side "not found" screen. Crawlers treat that as a real page.
    • Navigation that is not crawlable. Buttons with click handlers, instead of real href links, hide the rest of the site from a basic crawl.
    A laptop on a desk showing a finished looking website next to a nearly empty HTML document

    A worked example

    Imagine a local booking tool shipped on Lovable. The home, pricing and login pages work in Chrome. curl on the homepage returns a short document whose title is the app name, and that same title appears on every route. /sitemap.xml 404s. /robots.txt is missing. A typo like /pricng returns 200 with a "page not found" view. The header uses buttons, so a crawler that does not run those clicks cannot reach pricing.

    Paste the live URL, the builder name, and the routes that should rank into the vibe-coded site SEO prompt. Give it the repo if you have one, and ask it to inspect before it edits. You should get a findings table plus ready-to-paste files: a robots.txt that points at the sitemap, a sitemap that lists only canonical 200 URLs, unique titles and descriptions, Open Graph tags, and Organization and WebSite JSON-LD with no invented ratings or prices. It should also tell you how to return a real 404 for unknown paths.

    If the app is client-rendered, the output should name a prerender or server-rendering path that fits that builder. Adding meta tags in a React effect is not enough if they never appear in the first HTML response. After the patches, curl the homepage and a couple of inner routes and check that the unique titles, canonicals and JSON-LD are present in the raw HTML you get from curl.

    How to run it

    1. Copy the Fix SEO on a vibe-coded site prompt and paste it into Claude, ChatGPT, Cursor or whatever you use to edit the repo.
    2. Fill in the placeholders: the live URL, the builder, the framework if you know it, the routes that should rank, the brand title, and whether you have Search Console access.
    3. Let it inspect the codebase or the live HTML first. Then apply the patches it produces, one area at a time, and keep anything that already works.
    4. Verify with curl, by opening /sitemap.xml and /robots.txt, and by requesting a made-up path and checking the status code. If you have Search Console, verify the property, submit the sitemap, and use URL Inspection to confirm the rendered HTML now includes your titles and body copy.
    A tidy desk with a printed sitemap, a robots file and a browser open on a blank search console style screen

    Once the basics are in place

    This prompt is for the SPA failures that stop a newly built site getting crawled at all. If you already have crawl data or Search Console coverage issues across a larger site, use the technical SEO audit checklist and work from the export. If the problem is mobile rendering, Core Web Vitals or content parity between phone and desktop, use the mobile-first SEO checklist. Those two assume a site that can already be crawled. This one gets a vibe-coded app to that point.

    Search Console still needs a person. No prompt can click the verification check or submit the sitemap for you. Once the HTML, sitemap and status codes are right, that last step is what tells Google the routes exist.