Question explained · gr.AI

Why my competitors show up in AI answers and I do not

Because AI engines do not choose between you and your competitor. They choose among the businesses they can read. If they have a site that spells out what they do, where they are and what hours they keep, and you only have a Facebook page, they appear in the answer, get mentioned by name, and you do not. It is not preference, it is availability.

We measured how often this happens. In a single sector, across 7 Romanian towns, we found 493 businesses: 74% had no site at all, and of the 126 that did, 84% declared nothing about themselves in a format a machine can read. The average score of those with a site: 60 out of 100.

Three causes, in the order we meet them

1. You do not exist in writing

By far the most common: 74% of the businesses we found had no site at all. A Google listing and a Facebook page tell the engine that you exist, but almost nothing about what you do. An assistant that has to recommend a business to someone needs a sentence it can quote, and that sentence has to live somewhere.

2. You exist, but you do not say who you are

84% of the sites we scanned declared no structured data, meaning they said nowhere, in the format machines read, what they are called, what they are, where they are and how to reach them. The engine has to guess from the text. Sometimes it guesses well. The competitor who declares it explicitly does not depend on guessing, and 59% did not even have an llms.txt saying what matters on the site.

3. You are blocked

The rarest, 2% in this measurement, and the most serious, because you are not invisible, you are forbidden. Usually it is a forgotten line in robots.txt. But we also met the version that is completely invisible: a site where robots.txt explicitly allowed AI crawlers while the firewall returned an error on every uncached page. In a browser it looked perfect. For two engines out of four, the entire catalogue did not exist.

What does not explain the difference

It is not a punishment.
The engine did not exclude you. It found nothing to quote about you.
It is not about company size.
In our measurement the average score was 60 out of 100, so the whole sector is halfway there. The competitor who shows up is not necessarily bigger, they are more legible.
More content does not fix it.
Twenty articles on a site the AI crawler cannot enter stay twenty unread articles. Access gets solved before content.

What you can check yourself, in ten minutes

Open your robots.txt

Type your site's address followed by /robots.txt and look for four names: GPTBot (ChatGPT), ClaudeBot, PerplexityBot and Google-Extended (Google's summaries). If any of them appears with Disallow: /, you have found the cause without looking any further.

Test as a robot, not as a human

A firewall can let the browser through and block the bot, and from a browser you never see it. Check on an inner page, not the front page, because the front page is usually cached and answers anyone:

curl -A "GPTBot" -I https://site-ul-tau.ro/o-pagina-interioara
curl -A "ClaudeBot" -I https://site-ul-tau.ro/o-pagina-interioara

You want 200. Any 4xx or 5xx there, while the page opens fine in a browser, means you are blocked at the network level.

Look at what you declare

Run the page through Google's rich results test or the schema.org validator. The simple question: does a block appear saying what kind of business you are, with name, address and phone? If nothing appears, you are in the percentage above.

Ask the engines

Ask the question your customer would ask, not your company name but their need, on ChatGPT, Gemini, Claude and Perplexity, several times each. Note each time whether you appear, in what position, and whether you are cited with a link. Repetition matters: answers vary, and a single attempt tells you nothing.

The methodology and its limits

The businesses come from OpenStreetMap, scanned in 2026-08-05. The sector covered so far, named as it appears in the Romanian data: cabinete stomatologice. On every site found we checked structured data, the presence of llms.txt, AI crawler access by user-agent, and HTTPS.

What it does not measure: it says nothing about the quality of a company's services and does not guarantee that the company appears or does not appear in any given answer. It measures whether it can appear. And not all towns weigh the same: in some we found only a handful of sites, so the percentage there is illustrative, not statistical. The per-town figures, with the number of sites behind each, are published in full.