AI-Built Website
Launch Checklist
An AI can build a website in an afternoon. These ten checks tell you whether it's ready to be found and to bring in customers before you launch it, or hand it to a client. Each takes a few minutes, and none needs special tools.
How we tested. Checks 1 to 6 were tested on a small React site we built for this, on an exported and published Lovable project and Replit project, and on live AI-built sites. The others are standard practice, and we haven't tested them on any one builder. Last checked 16 October 2026.
1. Your content is in the HTML
Check: open each important page (homepage, services or menu, contact) and choose View Page Source. Search for a full sentence from the page. If it's not there, a crawler that doesn't run JavaScript sees an empty page. The full method is in the 10-second check.
Fix: pre-render or server-render the pages, so the content is in the HTML the server sends. For a React and Vite site, follow the React and Vite guide.
2. Every page has its own title and description
Check: view the source of three different pages. The <title> and <meta name="description">should differ on each and describe that page. In our check, several Lovable pages used a project name as the title, or “Lovable Generated Project” as the description.
Fix: set a title and description per page in the code. Keep titles to about 60 characters and descriptions to a sentence or two.
3. A made-up URL returns a 404
Check:run this with a page that doesn't exist. The first line of the answer should say 404.
curl -I https://yoursite.com/no-such-page
In our check, the Replit sites and the Lovable template previews we tried returned 200for a made-up URL (a “soft 404”), while a new Lovable project and a site built with Next.js returned 404. A 200 on a page that doesn't exist tells crawlers that every address is a real page.
Fix: add a real not-found page and check how your host handles unknown paths. In our test, a plain 404.html in the public folder returned a 404 on a static file server. Hosts differ, so test yours.
4. sitemap.xml and robots.txt exist
Check: open /sitemap.xml and /robots.txt. Look at what comes back: they should be a list of your pages and a few rules, not your homepage. Single-page apps often return the homepage for any address, so a missing file can look like it exists. On the Replit project we published, /sitemap.xml and /llms.txt both returned 200, but the response was the homepage (content type text/html), not a real file. In our check of nine sites, none had a sitemap.xml, and only four had a robots.txt.
Fix: put both files in the public folder. List your real pages in the sitemap, and point to it from robots.txt:
User-agent: * Allow: / Sitemap: https://yoursite.com/sitemap.xml
5. Business details are real text, with structured data
Check: your name, address, phone number and opening hours should appear in the page source as text, not only in an image or a map. Then search the source for application/ld+jsonand check it includes your business type, address and hours. You can validate it with Google's Rich Results Test.
Fix: add LocalBusiness structured data (or a more specific type such as Restaurant) to the page head, with the same details as the visible text. In our test it survived pre-rendering, so it was in the raw HTML.
6. A real share image and title
Check: paste your link into a chat app or social post and look at the preview, or search the source for og:image. In our check, four of five Lovable sites still used Lovable's default share image.
Fix: set og:title, og:description and og:image on each page, with an image that shows your business.
7. Nothing from development is left blocking the site
Check: search the page source for noindex, and check that /robots.txt doesn't contain Disallow: /for everyone. Make sure you're launching on your own domain over HTTPS, and that the site doesn't sit behind a login.
Fix: remove anything left over from development. A noindex tag tells search engines to keep the page out.
8. Search Console is set up
Check: add the site to Google Search Console and verify it. Submit your sitemap. Then use URL Inspection on your homepage and look at the page Google fetched, to see what Google actually saw.
Fix: if the fetched page is empty, go back to check 1.
9. Analytics and conversion tracking work
Check: install Google Analytics 4 (or your tool) and track the action that matters: a booking click, a call tap or a form submission. Perform the action yourself and look for it in the real-time report.
Fix: embedded booking widgets can hide the action from your analytics. In the Arros QD case study, booking-intent clicks could be measured, but confirmed reservations couldn't. Decide up front what you'll count, and test it.
10. Your Google Business Profile matches the site
Check: for a business customers visit, claim the profile and make sure the name, address, phone number and opening hours match the website exactly, and that it links to the site.
Fix: change whichever is out of date. Mismatches between listings and the website make it harder to tell which details are right.
Frequently Asked Questions
Which checks matter most?
Checks 1 to 4. They decide whether crawlers can read the site at all: the content is in the HTML, pages have their own titles, made-up URLs return a 404, and the sitemap and robots.txt work. Do those first.
Is this only for AI-built websites?
No, the checks apply to any website. We wrote it for AI-built sites because in our small check most of these basics were missing from them: none of nine had a sitemap.xml, and six of nine served an empty page to a crawler.
Will passing all ten get me recommended by ChatGPT?
No. These checks make a site readable and trackable. Whether an AI recommends a business also depends on reviews, listings and mentions elsewhere on the web.
Sources
Check the
technical side
Chad's free scan checks how crawlers read your site, and gives every technical issue with the fix.
Run the free scan →