Building your site with AI won't make AI recommend you.

ChatGPT and Gemini recommend businesses based on what they can read: the Google Maps listing, the reviews and a site their crawlers understand. A template site fixes neither that nor the design.

By @ignapradosSEO and AI4 min read

More and more dentists, salons, repair shops and restaurants build their own site with Lovable, v0 or Bolt in an afternoon. What comes out looks like everyone else's: the same centered hero, the same icon cards, the same WhatsApp button. Fine, and identical to hundreds of others. On top of that, the business doesn't come up when someone asks ChatGPT for "a dentist open on Saturday near me".

That's two problems. The design one is visible at a glance. The other isn't: AI doesn't look at a site the way a person does. It reads the HTML, checks it against Google Maps and other sites, and recommends whoever it can verify. AI-built sites tend to fail in that part too, the one nobody sees.

Where AI gets its recommendations

Google has it in writing. Its page on AI features and your website says AI Overviews and AI Mode rely on the regular index: a page that isn't indexed can't be cited, and the basics it lists include keeping your Business Profile up to date.

On Google Maps, local ranking comes down to three things: relevance, distance and prominence. Prominence draws on, among other signals, how many websites link to the business and how many reviews it has.

ChatGPT searches with its own crawler, OAI-SearchBot. According to OpenAI's documentation, a site that blocks it won't show up in ChatGPT's search answers.

What AI-built sites tend to miss

  • The text arrives empty. Lovable, Vite and Bolt build the page with JavaScript in the browser. AI crawlers read the HTML the server sends, and if the address and opening hours aren't in it, they don't exist.
  • The crawlers are blocked. A copied robots.txt or a firewall that stops unknown bots shuts out OAI-SearchBot without anyone noticing.
  • It never says what or where it is. Without local business structured data, a machine has to guess that "Bright Smile" is a dental office on a particular street.
  • The details don't match. The site says one set of hours and the Maps listing says another. Google asks for accurate information, and a profile with wrong details may not show up for the searches it should.

The markup it's missing

The most concrete fix is a LocalBusiness structured data block in the HTML. Google requires at least name and address, recommends telephone, url, geo and openingHoursSpecification, and asks for the most specific subtype available (Dentist, Restaurant, HairSalon). It goes in the <head> of the home page:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Dentist",
  "name": "Bright Smile Dental",
  "url": "https://brightsmiledental.com/",
  "telephone": "+1 512 555 0142",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "1200 South Congress Ave",
    "addressLocality": "Austin",
    "addressRegion": "TX",
    "postalCode": "78704",
    "addressCountry": "US"
  },
  "geo": { "@type": "GeoCoordinates", "latitude": 30.25012, "longitude": -97.74921 },
  "openingHoursSpecification": [{
    "@type": "OpeningHoursSpecification",
    "dayOfWeek": ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday"],
    "opens": "09:00",
    "closes": "18:00"
  }]
}
</script>

The details in the example are made up: use yours, exactly as they appear on your Google Maps listing. And don't treat it as magic: Google says there's no special structured data needed to appear in its AI answers. What it does is let any machine read, without guessing, who you are, where you are and when you're open.

Your Google Maps listing, and a designer's hand

For a local business, the website is half the job. The other half is the listing: the right category, current hours, photos of the place, recent reviews with replies, and your own site in the "Website" field. That's what Gemini and Google Maps read first.

Both are the work of Alavista, Igna's other project: websites for local businesses in Latin America, built so Google and AI assistants recommend them, with the Google Maps listing put in order. They're designed by Igna, who has been designing websites for more than six years, so they don't come out with the template look AI leaves on its own. Its free check of whether AI can recommend your business finds your listing and tells you what's missing (the site is in Spanish).

How to check it today

  1. Look up your business on Google Maps and compare the name, address, phone number and hours with your site. They should match character for character.
  2. Open your site with Ctrl+U (view source) and search for your address. If it isn't in that text, crawlers can't see it.
  3. Open `yourdomain.com/robots.txt` and make sure there's no Disallow for OAI-SearchBot or for every bot.
  4. Run your site through the [Slop Score](@slopometro). The free check measures your home page; Pro goes through every page and flags, among other things, "Without JavaScript, your page is blank", "You're blocking the AI tools that search" and "It has no structured data".

If step two or three came out wrong, there is a post for each: how to test with curl whether ChatGPT can read your site and what to put in robots.txt so you stay in ChatGPT. With a readable site and an up-to-date listing, AI has somewhere to find your name.

Sources

Frequently asked questions

Why doesn't ChatGPT recommend my business?

Usually because it can't find details it can verify: the site blocks its crawler or arrives empty, the Google Maps listing is incomplete, or the details don't match from one place to another.

Do I need an llms.txt for AI to recommend me?

Not for Google: it says no new files are needed for its AI features. What you do need is for the page to be indexed and for crawlers to be able to read the text.

What matters more, my website or my Google Maps listing?

For a local business, the listing. The website helps when its details match the listing and crawlers can read it.