Vibe code SEO
SEO for websites built with Lovable, Cursor, v0, Bolt, and Replit. Your site is not ranking badly. It is not being read at all. I find the exact reason in the raw response, fix it inside your codebase, make Google come back and look again, and then do the ranking work your site could not use before.
Why the fix you already made did nothing
Three separate clocks, and the tools only show you the first one. Most vibe coded businesses that try to fix this themselves get the build right and then wait months for nothing, because nobody told them the second clock has to be started by hand.
Starts when you fix the code. Stops the same day. Server rendering, canonicals, and a real sitemap can be corrected in an afternoon.
Fixed in hours
Starts only when Google re-fetches the corrected page. On a young domain that has already been filed as duplicates, it does not start on its own. Somebody has to request it, page by page.
Forced, then weeks
Starts once the page is indexed and eligible. This is where ordinary SEO finally applies: the query map, the page copy, the links, the profile. It cannot start before the other two.
Months, and it compounds
What the x-ray checks
These come from real sites I have diagnosed, not from a checklist I copied. Every one is read from the raw server response, which is the only thing search engines see.
<link rel="canonical" href="https://yourbusiness.com/" />
<body><div id="root"></div></body>
GET /this-page-does-not-exist -> 200
<url><loc>…</loc><priority>0.8</priority></url>
document.querySelectorAll("script[type=application/ld+json]").length // 0 in source
URL Inspection -> Request indexing (per page)
How the engagement runs
The order is the point. Ranking work on a page Google has not read is wasted money, and most of what vibe coded businesses have already paid for was exactly that.
I fetch your site the way Googlebot and the AI crawlers do, byte for byte, and compare it with what a browser shows. You get a written report of what the crawler actually received, with the evidence attached. Free.
Canonicals, render mode, real 404s, sitemap, per-page titles and schema in the server response. Done in your codebase, in your repo, without touching the design you built.
The fix is invisible until Google looks again. I request indexing page by page, change the lastmod signals, and watch Search Console until the pages move from excluded to indexed.
Now the site can use normal SEO. A keyword-to-page map from real search data, page copy that answers the query, internal links, and a Google Business Profile if you serve a local area.
Proof, not promises
The first two are new sites taken from launch to their first real traffic, one of them designed, built, and written with an AI-native workflow. The other two are what the rank clock looks like once it has been running for a while.
481
clicks from launch · 3 months
A new tech site to 481 clicks and a climbing curve in 90 days
312
clicks in 28 days · First month
312 clicks in the first 28 days of SEO
10.8K
organic clicks · 3 months
10.8K clicks from 1.86M impressions for a Queens dermatology clinic
23.3K
organic clicks · 6 months
23.3K clicks from 3.55M impressions in six months
All numbers are taken directly from Google Search Console, Google Business Profile, GA4, Bing Webmaster Tools, and the Microsoft AI Performance dashboard for active clients and owned projects. Names withheld at client request.
Every builder, the same job
The tool changes where the failure sits. It does not change what has to be true for a page to be read: the words in the server response, one honest canonical per URL, real 404s, and a sitemap Google can trust.
Lovable
React and Vite SPA by default; SSR is a setting, not the default.
Cursor
Whatever the prompt produced. Usually Next or Vite, usually client-only.
v0
Next.js output that still needs metadata, canonicals, and a sitemap wired in.
Bolt
Vite SPA with a shared index.html for every route until told otherwise.
Replit
Anything. The hosting layer is where the 200-for-everything problem lives.
Framer and Webflow AI
Rendered HTML, so the failures move to duplicate pages and thin templates.
Want the site built right the first time instead? That is the design service. Want to be the name AI gives? That is AI search.
Questions
SEO for websites that were built by describing them to an AI tool, such as Lovable, Cursor, v0, Bolt, or Replit. Those sites fail search in a specific, predictable way that ordinary SEO audits do not catch, because the audit tools read the rendered page and the search engines read the raw response. Vibe code SEO starts from the raw response, fixes what the crawler receives, forces the recrawl, and only then does the ranking work.
Yes, and it is the most common version of it. The homepage indexes because its canonical points at itself. Every other page carries the same canonical, so Google files them as copies of the homepage. Your site is indexed as one URL.
Almost never. The fixes live in the document head, the render mode, the routing, and the sitemap. I make them inside your existing codebase, on a branch you can review, and your design stays exactly as you built it. If yours is the rare build that genuinely cannot be fixed in place, I will say so before any work starts.
Because fixing the build and fixing the index are two separate clocks. Google recorded its duplicate decision and stopped coming back. Nothing forces it to re-read the corrected pages until someone requests it, page by page, and gives it a fresh lastmod signal. That step is part of every engagement, and it is the one most people skip.
No, and I will not pretend otherwise. Google's Indexing API officially covers job postings and live broadcast events only, and Google does not use IndexNow. The honest route is per-URL indexing requests in Search Console plus correct lastmod dates, and then patience measured in weeks, not months.
Indexing usually moves within a few weeks of the fix plus the recrawl requests, and I show you the Search Console coverage report as it changes. Ranking depends on your market and your authority, the same as any site. What I can promise is that the site will finally be eligible, which it was not before.
Your URL, access to your repository or builder project, and Search Console. If Search Console is not set up yet, that is the first thing we do together, because every step after it is measured there.
Work with Ashikur Rahman
Send me the URL. You get a written x-ray of the raw response, the exact reasons your pages are not indexed, and what I would do about each one. Free, and I read every one myself.
