Vibe Coding
SEO

Crawlability

A lot of basic SEO is just markup an AI can add for you. The steps I followed to make a vibe-coded site a little easier to find — in plain English.

CA Gaurav K PatiyatBlog #3~5 min read
Flat illustration: two people at a desk building a website — code on the left of the browser window, a page layout with an image and text on the right.
In this post
On-page SEO checklist (no code)SEO for AI-built & no-code sitesRanking in AI search & AI OverviewsA copy-paste SEO audit prompt

When a website can’t be found in search, it almost doesn’t matter how good it is. Making sure search engines can find and understand a site is what SEO — search engine optimisation — is for.

It has a reputation for being technical and a little dark-arts, but most of what matters is surprisingly ordinary: clear titles, a short description for each page, a few tags that tell search engines what they’re looking at, and a simple map of the site. It’s mostly just markup — and markup happens to be something AI is good at. You can ask an assistant to add nearly all of it for you, and keep your own attention on what the page actually says.

Search engines can’t send anyone to a page they don’t know exists.

So, the same way I’d added security layers, I worked through the search side. Here are the five steps that seemed to help most, in plain English — plus a sixth for the way AI is starting to change search.

Step 01

Titles & descriptions — what search shows

The clickable headline and the grey summary line under it. Every page gets its own, written around the words people really type. Leave it out, and the search engine guesses for you — usually less well.

Step 02

Structured data — labels machines can read

Invisible tags (in a format called JSON-LD) that spell out what a page is — a person, an article, an FAQ. It helps a search engine, or an AI answer, tell who wrote what — and is a bit more likely to get it right.

Step 03

A sitemap & a robots file — a map for crawlers

A plain list of every page (the sitemap), and a short note saying “you’re welcome to look around” (robots.txt). Together they hand search engines a map instead of making them guess their way through.

Step 04

Real pages, not in-app routes

A lot of AI-built sites keep everything behind one web address with a # in it. Search engines can’t file those away as separate pages, and shared links preview badly. Giving each post its own real, permanent URL probably made the biggest difference for a small site like mine.

Step 05

Link previews — how it looks when shared

The title, blurb and image that appear when someone drops your link into WhatsApp or LinkedIn. A few “Open Graph” tags and one 1200×630 image decide whether a shared link looks deliberate, or like a broken scrap.

BonusOne more layer
Step 06

Writing for AI answers, not just blue links

Search is changing: a search engine now answers a lot of questions itself, right at the top, and tools like ChatGPT do too. They quote pages that are easy to lift from — so I open each post with a two-or-three-sentence direct answer, keep paragraphs short, and add a small FAQ marked up for machines. It feels a bit like writing a good executive summary: lead with the answer, then back it up.

None of this made me an SEO expert, and it doesn’t need to. Most of it was fairly unglamorous setup — much of which an AI can handle — followed by the slower, harder part: publishing things worth reading, and giving search engines time to notice. Build it, secure it, then try to make sure it can be found.

Use this prompt

No coding required. Paste this into your AI tool to add the same steps to your own site:

You're an SEO expert. Add a complete on-page SEO layer to the website I built with AI, without changing how it looks or behaves. Keep everything copy-paste ready. FIRST, scan the site and tell me: every page or route, whether pages have real URLs or #-style routes, and what's already there (titles, meta, sitemap, structured data). Then add or fix: (1) a unique, keyword-aware title and ~150-character description on every page; (2) one main heading per page and a clean heading outline; semantic HTML and a page language; (3) Open Graph + Twitter Card tags on every page, pointing to a 1200x630 share image; (4) a self-referencing canonical link on every page; (5) JSON-LD structured data: Person or Organisation sitewide, Article on each post, FAQ where there's a Q&A; (6) descriptive alt text on every image, and readable hyphenated URLs; (7) a sitemap.xml listing all pages, and a robots.txt that allows crawling and points to it; (8) if pages use #-routes, generate real separate pages, each with its own title, description and canonical, so they can be indexed individually; (9) a 2-3 sentence direct answer at the top of each page, plus a short FAQ, so AI search can quote it. RULES: don't break the design. Confirm each change as you make it. At the end, give me a SEPARATE checklist of the things only I can do in a dashboard — connect a domain with HTTPS, verify the site in Google Search Console, submit the sitemap, and request indexing — with click-by-click steps and the exact values I'll need.