01
Words in the server response
Your page text, headings, and answers arrive in the first HTML, before any JavaScript. On a vibe coded site this is the fix that comes first, because without it every other step has nothing to work with.
Vibe code AI search
Be the business ChatGPT recommends.
AI search optimization for websites built with Lovable, Cursor, v0, and Bolt. Your customers now ask an assistant who to call before they ever open Google. If your site handed the crawler an empty page, you are not in that answer. I get you into it, and I show you the citation when it lands.
How an assistant decides who to name
Four steps happen before the answer, and a vibe coded site fails the first one.
None of this is mysterious. The model is not judging your design. It is retrieving records, and it can only retrieve what it was able to collect.
- Fails here
Fetch
A crawler such as GPTBot, ClaudeBot, PerplexityBot, or Google-Extended requests your page. If robots.txt blocks it, or the page is a blank root div, this is where it ends.
- 02
Extract
The text, headings, and schema in the raw response are pulled out. Nothing that only appears after JavaScript runs is part of this step for most engines.
- 03
Store
What was extracted is indexed alongside everything the web says about you: directories, reviews, forums, news. The engine now has a record of your business, or it does not.
- 04
Cite
When someone asks, the model retrieves the records that answer the question and names the businesses in them. You are in the answer only if you were in the record.
The engines also read what the rest of the web says about you, and they lean on it heavily. That is why the work has an off-site half, and why it is not just content.
Start with the indexing fixWhat has to be true
Six things an engine needs before it can name you.
Every one of these is checked in the free x-ray. Most vibe coded sites fail the first two outright, and the rest cannot help until those are fixed.
02
The crawlers let in, by name
GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, and the Bing crawler explicitly allowed in robots.txt, plus an llms.txt that tells them which pages matter. This site does both. Most vibe coded sites do neither.
03
One consistent entity
Organization, Person, and Service schema in the raw HTML, with the same name, address, phone, and services as every directory that lists you. Engines verify a business by matching records. Mismatches get you dropped.
04
Answer-shaped pages
Pages built around the questions people actually ask an assistant, with the direct answer high on the page and the facts an engine can lift: hours, prices, areas, who it is for, who it is not for.
05
The off-site record
Engines cite directories, review sites, and forums far more than they cite the business itself. I map which platforms an engine is already citing for your queries and get you accurately listed there. Free platforms only.
06
Measured where it shows up
AI referrals in GA4, the AI Performance report in Bing Webmaster Tools, and a fixed set of prompts re-run every month so you can see the name appear. If it cannot be measured, I do not claim it.
Proof, not promises
Citations you can count, in someone else’s dashboard.
The AI Performance report in Bing Webmaster Tools counts every time Copilot and its partners cite a page, and the Bing search report shows the clicks that follow. None of these are numbers I typed.
Microsoft AI Performance dashboard
208.4K
AI citations · 3 months
208.4K AI citations in three months
Microsoft AI Performance dashboard
184.8K
AI citations · 3 months
184.8K AI citations for a content site in three months
Bing Webmaster Tools
3.1K
Bing clicks · 3 months
3.1K Bing clicks feeding Copilot answers
Bing Webmaster Tools
2.6K
Bing clicks · 3 months
2.6K Bing clicks the competition ignores
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.
How the engagement runs
Five steps, and the first one is a baseline you can hold me to.
I record what the engines say before any work starts, so the result is a before-and-after you can read yourself, not a report I wrote about myself.
- 01
What the engines say now
I ask ChatGPT, Perplexity, Copilot, and Google AI the questions your customers ask, with and without your name, and record who gets named and which sources are cited. This is the baseline everything is measured against.
- 02
Fix the read
Server-side text, crawler access, llms.txt, and per-page entity schema in the raw response. On a vibe coded site this is most of the gap, and it is the part nobody else checks.
- 03
Shape the answers
Rewrite or add the pages that match the questions from step one, structured so an engine can lift the answer and the facts around it. Your voice, your facts, no filler.
- 04
Earn the mentions
Get you listed, accurately and consistently, on the free platforms the engines are already citing for your market. Not link building. Record building.
- 05
Measure and repeat
The same prompts re-run monthly, AI referrals tracked in GA4, citations tracked in Bing. You see the name show up, or you see exactly why it has not yet.
Questions
What vibe coded businesses ask about AI search.
What is vibe code AI search?
Getting a website that was built with an AI tool, such as Lovable, Cursor, v0, or Bolt, to be read, cited, and recommended by AI assistants such as ChatGPT, Perplexity, Copilot, and Google AI Overviews. It is the same discipline people call answer engine optimization or generative engine optimization, applied to sites that usually fail the very first step, which is being readable by the crawler at all.
Why does ChatGPT not know my business exists?
Because it never received anything about you. Either the crawler was not allowed in, or it was allowed in and got an empty page, or the only places that mention you online are inconsistent with each other. On a vibe coded site it is usually the second one. The fix starts with the raw HTML, not with content.
Is this different from SEO?
It shares the foundation: a readable site, honest structure, and a consistent record of who you are. It differs in what it targets. SEO earns a position on a results page. AI search earns a mention inside an answer, and the engines lean far more on third-party sources for that than Google does for rankings. Most of my clients do both, because the fixes overlap.
Do I need an llms.txt file?
It is cheap, it is harmless, and this site has one. It is also not the thing standing between you and AI visibility. I add it as part of the crawler-access work, after the page text is actually in the server response. Anyone selling llms.txt as the fix on its own is selling you the smallest part.
How do you prove it worked?
Three ways. The prompts from the first month get re-run and recorded, so you can read the answer with your name in it. Bing Webmaster Tools has an AI Performance report that counts citations directly, which is where the 208,400 figure on this page comes from. And Search Console shows the clicks that follow once the pages are readable.
How long does it take to get cited?
It varies by how much of the web already mentions you. A site whose pages are readable and answer-shaped from day one can be cited within weeks. A business with no off-site record takes longer, because that record has to be built. I give you a timeline after the baseline, not before.
Work with Ashikur Rahman
Find out what AI engines already say about you.
Send me the URL. I will check whether the AI crawlers can read your site at all, ask the engines about your business, and send back what I find. Free, and a real person reads it.