Skip to content
← Field notes

Founder measurement · Thailand · · 6 min read

We asked the models to find Thailand’s dive operators. 11 of 16 came back.

We hand-assembled a set of 16 real Thai dive operators — each with its own website, no aggregators and no directory profiles — and ran it through the ordinary pipeline. 11 came back with a citation we can stand behind. 5 did not.

How it was counted

The question was asked about the category, with no company name inside it. That matters: a query containing the name makes the platform go looking for that company, find its site and cite it, after which almost any business with a working website gets a mention.

A mention was counted only on a real citation with a source. A name matching in the text is not enough — models repeat the name from the question in their own preamble, and counting that means measuring the echo of our own query.

The query is in English. That is a measured result, not an assumption: for Thailand the English query returned twice as many real operators as the Thai one. Thai dive centres sit on .com rather than .th because they sell to foreigners.

Who was found

  • Discovery Divers
  • Sea Bees Diving
  • Silent Divers
  • Assava Dive Resort
  • Big Blue Diving
  • Pattaya Dive Centre
  • Oceanic Dive Center
  • Scuba Shack Koh Tao
  • Sea Dragon Dive Center
  • The Life Aquatic
  • Koh Tao Divers

What this result does not mean

We do not name the 5 operators that were missed. Publishing a list of “these real businesses are invisible to AI” on our own marketing site is a negative claim about named third parties that the finding does not require. The number carries the argument; the names would carry only collateral damage.

The set of 16 operators was assembled by hand and is not a random sample of the Thai market. It does not follow from this data that “31% of Thai dive centres are invisible” — only that in this particular set of 16, 11 were found.

This is one run, one day and one phrasing. On our data, different phrasings of the same question give different answers, and the same question on different days does too. A single-day reading shows direction, not magnitude.

How everything above is counted — methodology.