Five doors in Osaka, twenty in Munich: how many directories actually lead to you?
24 August: assistants do not read the same pages in every language. Today: they do not read the same NUMBER of them. In Busan, one domain is cited for all eleven trades.
11/11
the trades blog.naver.com is cited for in Busan - every one
In this article
Short answer: the number of entry points leading to you varies fourfold by market, and that changes what a listing is worth. In Osaka, only five domains are cited across several trades. In Munich, twenty. And in Busan, a single domain is cited for all eleven trades measured.
5 to 20
the number of doors by market, on an identical question
18/21
market pairs that share no door at all
On 24 August we showed that an assistant does not read the same pages depending on the language of the question. That was a measurement on software sectors. This one covers eleven local trades - baker, plumber, lawyer - across seven cities, and asks a question the earlier one did not: not *which* pages, but how many.
The question is not which directory to join. It is how many there are where you trade - because where there is only one, being absent from it is being invisible.
What we measured
On 28 August: 154 draws, eleven trades, seven languages, no failures. One city per language, the one where the trade is actually searched for: Valencia, Munich, Kraków, Osaka, Busan, Chengdu, Dubai. Two assistants that search, Perplexity and Gemini.
The criterion is structural, not semantic. A domain cited for at least three of the eleven trades, in one language, cannot be a single business's own website: no bakery is also a law firm. That is what we call a door. The threshold of three is a choice, it is written down here, and our script also outputs the domains at two trades so the choice can be judged rather than believed.
The number of doors varies fourfold
| Market | Domains cited | Doors | Leading domain |
|---|---|---|---|
| Munich (de) | 188 | 20 | werkenntdenbesten.de - 9/11 |
| Chengdu (zh) | 133 | 14 | 163.com - 5/11 |
| Kraków (pl) | 192 | 12 | pieknykrakow.pl - 8/11 |
| Valencia (es) | 212 | 11 | destaco.es - 6/11 |
| Busan (ko) | 132 | 8 | blog.naver.com - 11/11 |
| Dubai (ar) | 228 | 6 | bayut.com - 8/11 |
| Osaka (ja) | 172 | 5 | note.com - 4/11 |
The number of domains cited tells you nothing about the number of doors. Dubai cites the most - 228 - and has only six doors. Busan cites the fewest - 132 - and holds the most dominant domain in the whole measurement. A market can be talkative and concentrated, or sparing and dispersed: these are two independent properties, and it is the second one that decides your strategy.
In Busan there is one door
blog.naver.com is cited for eleven trades out of eleven. Not ten, not nine: all of them. No other domain, in any other market in this measurement, comes close - the next most dominant is werkenntdenbesten.de in Munich, at 9 trades out of 11.
For a business in Busan the consequence is blunt and simple: a presence on Naver is not part of the plan, it is the plan. For a business in Munich the opposite holds - twenty doors share the work, no single listing is decisive, and twenty careless listings are worth less than three kept up to date.
And these doors do not resemble each other across markets
Of the twenty-one possible pairs of markets, eighteen share no door at all. The three exceptions are thin: Valencia and Munich overlap by 3 %, Munich and Dubai by 4 %, Valencia and Dubai by 6 %. Polish, Japanese, Korean and Chinese share nothing with anyone.
Put plainly, the directory list a vendor sells you for France will do nothing for you in Osaka, and knowing that costs less than discovering it.
The limits
One city per language, one single day.
Valencia is not Spain and Osaka is not Japan. A measurement taken on 28 August does not say what these markets will do in October.
The three-trade threshold is a choice
, not a truth. It separates a directory from a business website without ever reading the content; domains at two trades exist and are listed by our script.
Being cited is not being chosen.
We count the domains the assistant reads to build its answer, not clicks and not customers.
Two engines, not three.
Perplexity and Gemini search the web at question time. ChatGPT's consumer surface is not measured here.
Nothing here proves causation.
That these domains are read does not demonstrate that appearing on them gets your business cited.
What you can do with it
Count the doors in your market before buying listings.
Where there are few, one listing changes everything; where there are twenty, none of them is enough on its own.
Check the leading domain in your market.
If one is cited for nearly every trade, your absence from it shows more than anywhere else.
Do not carry a list from one country to another.
Eighteen market pairs out of twenty-one share nothing.