97% are never named

Short answer: for 97% of local businesses the problem is not ranking low with an AI assistant. It is not existing anywhere in its answer. We audited 6,136 businesses across 18 markets, asking the questions their customers actually ask. 5,948 were never named. Not one mention, not one line, not one link. In Japan, 399 out of 399. In Saudi Arabia, 423 out of 423.

A French dentist's rival, inside the AI's answer, is not another dentist. It is Doctolib.

What we measured

For each business we ask AI assistants a buyer's questions - "who are the best X in Y?" - in the language of its market, never naming the business. Then we read the answer: is the business cited, in what position, and who is cited instead. As of 9 September 2026: 6,136 businesses, 18 markets - practices, clinics, agencies, shops.

The score out of 100 measures presence in those answers. Zero does not mean "ranked low": zero means absent from every answer.

Zero is not the exception, it is the rule

5,948 businesses out of 6,136 score zero, or 96.9%. The rate never drops below 92%, and three markets sit at 100% or one unit away.

And for the 188 that clear the bar, the win is narrow: their median score is 24 out of 100, and only 19 pass 50. Being named does not mean being recommended everywhere; it means being recommended one time in four.

Three names take everything

In a given city, for a given trade, we counted the businesses the assistant names. Across the 208 combinations where we audited at least five businesses, the top three names capture 100% of the citations, at the median. In 107 of them, three names take absolutely everything. Only three fall below half.

So this is not a ranking where you climb. It is a three-step podium, and the rest of the market is off-camera.

And on those steps, often, not a local business at all

We isolated the names cited in three or more cities of the same trade. By construction that is no longer a neighbourhood business: it is a chain, a platform or a professional body. They are only 237 names out of 3,983, or 6% - and they capture a quarter of all citations.

Look closely at the last two rows. Asked "which dentist in Lyon?", the name coming back most often is an appointment platform, not a dentist. Asked "which accountant in Liège?", it is the professional body. The assistant does not answer *who*, it answers *where to look*. It substitutes infrastructure for the answer.

What this measurement does not prove

• These 6,136 businesses come from our own campaigns: small local outfits, in 18 markets we chose. It is not a random draw of the world economy, and a larger business would probably score differently.
• We ask a limited number of questions per business. A business absent from our questions may well be named on a phrasing we did not ask.
• The score measures PRESENCE in an answer, not revenue. Nobody knows today what share of customers arrives through this channel.
• "Chain, platform or professional body" is a mechanical deduction - a name cited in three cities of the same trade is not a neighbourhood business - not a classification we made by hand.
• It is a snapshot. These systems change their sources and their answers within months.

What it changes for you

You can see in a minute, free and without an account, whether your business is named - and who is named instead.