Skip to content
Menu
Talk on WhatsAppPortuguês

AI and search

How to appear in ChatGPT, Perplexity and Google AI Overviews

Seven concrete site changes that raise the odds of your company being cited in AI answers, plus the two-minute test that shows whether crawlers can see your content at all.

8 min readFluxo Ads

More and more people ask ChatGPT what they used to type into Google. The answer that comes back cites two or three companies, and the rest do not exist in that conversation. This article lists what shifts that in your favour, in the order it usually pays off.

First: the two-minute test

Language model crawlers do not execute JavaScript. That means they see nothing that appears only after the browser assembles the page. To find out what they see of your site:

  1. Open your site in the browser.
  2. Disable JavaScript in site settings (in Chrome: the padlock beside the URL, Site settings, JavaScript, Block).
  3. Reload the page.

1. A direct answer under every heading

Right after the H1, write two to three sentences that answer the query on their own, independent of the rest of the page. State what it is, who it is for, what it includes and how it starts.

That is the passage models cite, because it is the only one that survives summarization intact. It comes before any button or call to action: a page opening with a CTA before content gives the model nothing to quote.

2. Tables instead of comparative prose

Models extract tables with far more fidelity than running text, because the relationship between columns is declared rather than inferred. Every comparison, price range, timeline or criteria list performs better as a table.

3. FAQs with structured data

A frequently asked questions section marked up with FAQPage delivers questions and answers already paired. Beyond earning rich results on Google, it is the easiest format to cite there is.

An implementation detail worth following: use the HTML `details` element rather than a JavaScript accordion. It works without scripts, is keyboard accessible by default, and collapsed content stays in the HTML.

4. Facts in full, never hidden behind a click

Phone, address, hours, turnaround and price ranges must be written into the text. "Click here to see the address" delivers no data to whoever is reading the HTML, and that is exactly how a company ends up cited with wrong information.

5. Correct structured data

Schema.org for organization, service, article and navigation states unambiguously what you do, where you operate and how to reach you. Verify it in Google Rich Results Test before calling it done: schema with errors is ignored entirely.

  • Organization: name, logo, contact and external profiles confirming the company exists.
  • Service: one per service page, pointing at the organization as provider.
  • Article: on posts, with publication date and author.
  • FAQPage: on frequently asked questions.
  • BreadcrumbList: so the model understands site hierarchy.

6. An llms.txt file

A plain-text file at the site root with a business summary and a map of the pages. The sitemap says which URLs exist; llms.txt says what each one answers and what the company facts are.

The standard is still consolidating and not every crawler consumes it today. But the cost of maintaining it is near zero when it is generated from the same registry that generates the sitemap, and the payoff, if and when adoption comes, is direct.

7. Presence beyond your own site

Models trust companies mentioned in several places. A complete Google Business Profile, an active LinkedIn, presence in industry directories and citations on other sites together form the signal set confirming the company is real.

This is also where AI optimization and traditional SEO meet: both reward the same thing, which is being a verifiable entity.

How to know it is working

Build a list of ten to twenty questions your customers would genuinely ask, run them all through each assistant, and record who gets cited. Repeat monthly. It is laborious and it is the only honest way to measure, because there is no rank tracker for AI answers.

In parallel, isolate sessions in analytics arriving from chatgpt.com, perplexity.ai and gemini.google.com. That number usually starts small and grows fast.

If you want us to do this work, it is the AI search optimization service. If you prefer to do it in house, the list above is the order we recommend.

Frequently asked questions

Does blocking AI crawlers protect my content?
Blocking prevents citation along with reading. For a media outlet that lives on visits, blocking makes sense. For a company that wants to be found and hired, blocking means disappearing from the conversation where the customer is deciding.
Do I need different content for AI and for Google?
No. What works for AI citation (clarity, structure, verifiable data) also works for traditional search. What changes is the priority of certain techniques, mainly structured data and JavaScript-free readable content.

More articles