Why Your AI-Built Site Is Invisible to Google
You used v0, Lovable, Bolt, Cursor, or another AI builder. The site looks polished. You shared it. Nobody found it on Google. This is not bad luck — it is four specific, fixable problems that AI tools produce by default and human architects catch before launch.
The Case Study Behind This Post
This site — paddock20.com — went from a Google Core Web Vitals score of 38 to 72 in a single audit sprint. The culprits were identical to every AI-built site we have diagnosed since: wrong canonical configuration, a sitemap pointing to stale routes, client-side rendering where Google needed static HTML, and topical identity drift from leftover placeholder content. Every fix is documented here.
The AI builder did not fail you. It built exactly what you asked for. The problem is that Google does not index websites based on how they look. It indexes them based on four technical signals that AI builders do not configure automatically — and that most founders never know to check.
The wave of AI-generated sites launched in 2025 and 2026 has created an unusual search landscape: thousands of beautiful, fast, functional websites that Google has essentially never seen. If yours is one of them, this is the diagnostic and the fix list.
Problem One: Google Cannot Read What the AI Built
Most AI builders default to client-side rendering. The page HTML that Googlebot receives when it crawls your site is a nearly empty shell — a few lines of React boilerplate, a bundle reference, and nothing else. The actual content — your headline, your services, your copy — only appears after JavaScript executes in the browser.
Googlebot runs a two-pass crawl: a fast HTML pass, and a slower JavaScript rendering pass that processes a small queue. Your site gets into the fast-pass queue and Googlebot sees nothing useful, so it deprioritizes the rendering queue slot. Result: your content never gets indexed, or gets indexed weeks late with incorrect signals.
The Fix
Server-side rendering (SSR) or static generation (SSG). If you built on Next.js via v0 or Vercel, your pages need to be Server Components or use generateStaticParamsfor static export. If you built on Lovable or Bolt (React SPA by default), the fastest fix is adding a prerendering service or migrating the money pages to Next.js. Check your rendering mode in Google Search Console → URL Inspection Tool → View Tested Page → HTML tab. If the HTML tab shows empty body content, you have a rendering problem.
Problem Two: No Sitemap — or a Sitemap Pointing to the Wrong Routes
AI builders often generate a sitemap.xml automatically. The problem is that the sitemap is generated at build time from whatever routes exist in the project — including demo routes, placeholder pages, admin scaffolding, and routes that no longer exist in the deployed version.
A sitemap with broken URLs tells Google your site is poorly maintained. Googlebot allocates crawl budget per domain — a sitemap full of 404s uses that budget on dead ends and leaves your real pages uncrawled. Worse, if your domain previously hosted different content (a Shopify apparel store, a different business), the old indexed URLs compete with your new sitemap for crawl budget.
The Fix
Audit your sitemap.xml manually. Every URL in it should return a 200 status. Remove anything that 404s, redirects, or points to a page that is not production-ready. In Next.js, use a programmatic sitemap.ts that only includes routes you explicitly list. Submit the corrected sitemap in Google Search Console and use the Removals tool to deindex any stale URLs from previous site versions.
Problem Three: Canonical URLs Are Missing or Wrong
A canonical URL tells Google which version of a page is the authoritative one. AI builders often ship without canonical tags, or with canonical tags that point to localhost, to a preview deployment URL (vercel.app), or to a URL pattern that does not match your production domain.
When Google sees content at both yoursite.vercel.app and www.yoursite.com with no canonical signal, it treats them as competing pages and either indexes the wrong one or splits ranking signals between both. Your production domain inherits zero authority from any content that was previewed on a Vercel deployment URL.
The Fix
Every page needs an alternates.canonical pointing to the exact production URL — https://www.yoursite.com/page, not a relative path or a deployment URL. In Next.js metadata, set alternates: { canonical: 'https://www.yoursite.com/page' } on every page. Block all preview and staging domains in Vercel project settings so Googlebot never crawls them.
Problem Four: Topical Identity Confusion
This is the one most builders never diagnose — and it is the one that creates the longest recovery time. AI tools generate placeholder content, demo copy, and example data. If any of it shipped to production — even a single page, even a footer link, even a metadata description that says "your company description here" — Google's topical classification of your domain is fractured.
Google classifies every domain by topic. If your domain was previously used for something else (a prior business, an apparel store, a parked domain), those old indexed pages dilute your topical signal for the new content. The site looks like it does not know what it is about. Rankings stall even when individual pages are technically well-optimized.
The other source of topical confusion: AI builders that add pages you did not ask for. Stock photography galleries. Demo contact forms. Example blog posts about unrelated industries. Every one of those sends a signal that your domain is not reliably about the thing you are trying to rank for.
The Fix
Crawl your own site with Screaming Frog or Ahrefs Site Audit (both free tiers are sufficient). Look for pages that do not fit your core topic. Either delete them (add 301 redirects to the nearest relevant page) or noindex them. Use Google Search Console Removals to force-deindex any URL that indexed with wrong content. Then publish three to five pieces of content that are unambiguously about your core topic — tightly scoped, with internal links connecting them to your main pages. Topical authority recovers faster when every new page reinforces the same signal.
The 30-Minute Audit Checklist
Before you do anything else, run through these four checks in Google Search Console. Each one takes under five minutes and tells you exactly which problem you have.
URL Inspection — Rendering Check
Search Console → URL Inspection → your homepage URL → View Tested Page → HTML tab. Is your main headline visible in the raw HTML? If not, you have a rendering problem.
Sitemaps — Coverage Report
Search Console → Sitemaps → submit sitemap.xml if missing. Then check Coverage → Excluded → "Page with redirect" and "Submitted URL not found (404)". Any count above zero is a problem.
Canonical Check
URL Inspection → any key page → check "Google-selected canonical" vs "User-declared canonical." If they do not match, or if Google-selected is a vercel.app URL, you have a canonical problem.
Coverage — Valid Pages
Search Console → Pages → Indexed. How many pages does Google think you have? If the number is higher than your actual page count, there are ghost pages from a prior site version still indexed.
What Agentic Engineering Gets Right From Day One
Everything in this post is a problem that does not exist when an architect — not just an AI tool — owns the build from the start. Server-side rendering is the default, not an afterthought. The sitemap is programmatic and maintained. Canonical URLs are set at the metadata level in the layout file. Topical identity is enforced through deliberate page architecture before any content is written.
The AI does not replace the judgment. It accelerates the person who already has it. That is the definition of agentic engineering — and it is the reason the sites we build show up in Google instead of waiting for a diagnosis post to find them.
If Your AI-Built Site Has These Problems
Run the free Align Diagnostic — it takes 90 seconds and shows exactly where your current setup is costing you. Or book a free 15-minute Fit Call and bring the audit results — we will tell you exactly what it takes to fix it and whether it is worth doing.
Build With RAIL: Free
90 Seconds to
Build Your App
With RAIL.
Answer three questions—type them or say them out loud—and watch RAIL scope your app in real time. Named app, working interface, and a build blueprint you keep. Before you ever talk to anyone.
A named app. A working interface. A blueprint PDF you keep. Whether we ever talk or not.