More and more people ask ChatGPT, Gemini or Claude "which dentist would you recommend in Estepona?" instead of searching Google. And more and more agencies sell the idea that they can "make the AI recommend you". This article explains what's actually true: what AEO (answer engine optimisation) and GEO (generative engine optimisation) are, what these models really look at before citing a website, and what nobody can promise you, however loudly the ad says it.
Crawlable isn't the same as citable
An AI being able to read your site is step one. Citing you is another. Models cite whoever:
- Answers concrete questions clearly: what you do, where, how much it costs, how to get in touch.
- Can be verified beyond their own site: reviews, directories, mentions in the press and on other websites.
- Is effortless to read: clean HTML, with the information visible without running JavaScript, and structured data that says "this is a business, here it is, this is what it offers".
The first is content. The second is reputation, and it's built off-site. The third is technical, and it's the only one a website can guarantee on its own.
The technical checklist you can control
What a website needs to be citable
HTML readable without JavaScript. Many AI crawlers don't run scripts. If your content is painted by JavaScript in the browser, to them the page is empty. A site served ready-made from the server, like the ones we build with Next.js, gets read in full.
Structured data. LocalBusiness with address, hours and phone; FAQPage for frequently asked questions; Article on the blog. It's how you tell the machine what each thing is.
Correct sitemap and robots. So they know which pages exist and you don't accidentally shut the door on them.
An llms.txt file. A recent convention: a text file at the root of your site summarising what you are and where to look. Cheap to add and increasingly read.
Pages that answer real questions. One page per service, an FAQ section, clear prices. Whatever a person would ask the AI, answered on your site.
Why so many sites fail the first box
The box most websites fail is the first one: readable without JavaScript. It happens with many visual builders, with themes that load content via script, and with single-page applications. The site looks fine in a browser and is blank to the crawler. You don't have to take our word for it — you can check. At app.wordnext.tech/comprobar you paste your address and, on top of the technical verdict, we return a report on how an AI would read you: what it understands about your business from your home page, what's missing, and three concrete improvements. Free and instant, no session with anyone required.
What we already publish (and why it matters)
AIs cite what they can cross-check. That's why we have public prices, an honest comparison of what we do well and what we don't, and a real case with figures: Olvidos went from 5.4 to 2.4 seconds of load time. We don't say it to show off: we say it because that's exactly the kind of content a model can cite without making anything up.
No fluff: nobody can guarantee you a place in an AI's answers. Not us, not whoever promises it. Models change, don't publish their criteria and mix sources. What can be done is to leave your site technically citable and audit it; the rest is built by your content and your reputation beyond your website.
What we do about it
Every site we deliver ships with server-rendered HTML, structured data, sitemap, robots and llms.txt. It's not an add-on: it's how they're built. And if you already have a site and want to know where it stands, the check tells you in five seconds. If you want to understand the speed side, which also counts, it's in is your website slow?.
Check now how an AI would read you. If the report leaves you with questions, the audit is free and we'll tell you, with numbers, what to change and what doesn't need touching.
Get every autopsy in your inbox
One article a week about websites that work (and the ones that don't): real costs, WordPress, speed and AI. No fluff; unsubscribe in one click.