Skip to content
Waseit
← All articles
AEOGEO

A clean website does not get you named

We crossed two measurements on 1,475 businesses: what their website tells a machine, and what AI assistants answer. Those that pass our three checks are named 2.3% of the time, the others 2.2%. The trade matters; the page title barely does.

2.3%

of sites passing all three checks are named

September 11, 20268 min read
In this article
  1. What we crossed
  2. The raw figure says yes, the country says no
  3. In the UK, where the names come out
  4. What really separates them: the trade, then the country
  5. What this measurement does not prove
  6. What we are changing in our own advice

Short answer: no, polishing your page is not enough to get named. Yesterday we recommended three free fixes so that a machine knows who you are and where. Today we checked whether they separate the businesses AI assistants name from the ones they ignore. Across 1,475 businesses, those that pass all three checks are named 2.3% of the time, the others 2.2%. Compare within the same country and the gap disappears. What makes the difference is elsewhere: a vet is named more than four times as often as average, an estate agent never.

2.2%

of the others are named too

9.4%

of vets are named, against 0% of estate agents

A readable page is a ticket in, not a seat. The question your customer asks has to be put to the answer itself.

What we crossed

Two measurements already existed, taken separately. The first, published on 10 September, reads a business website the way an answer engine would: three checks, two or three public HTTP requests. The second is our audit: we ask AI assistants a buyer's question, “who are the best X in Y?”, in the country's language, without ever naming the business, and we record whether it appears.

2,603 businesses have both measurements. We remove 1,128 of them: with no recognised trade, their audit asked a generic question (“a business like yours in Doha”), to which no assistant names anyone. Keeping them would have manufactured zeros. That leaves 1,475 businesses with a specific trade, in twelve countries, 33 of them named at least once: 2.2%.

Named

means cited at least once in the audit's answers, i.e. a score above zero. The median score of the 33 is 32 out of 100.

The three checks

: answer crawlers are allowed to read the site; a structured data block says who the business is; the page title names the business and the city.

Websites read

on 9 September 2026, audits run from 26 July to 9 September 2026.

The raw figure says yes, the country says no

Taken as a whole, two checks seem to pay. Businesses whose page shows a readable address are named 3.1% of the time, against 1.3% for the rest. Those that write their city in the title: 3.0% against 1.9%.

The effect does not survive one simple question: which country? The UK writes out addresses and cities more than almost any other market, and its businesses are also the most often named (4.8%). The Gulf and Japan rarely write their city, and their businesses are never named. Compare each business only with others in its own country, and the address advantage falls to +0.3 points, the city-in-title advantage to +0.4, and all three checks together to -0.6.

Share of businesses named by an AI assistant, by whether they pass each check. 1,475 businesses with a specific trade, twelve countries. “Same country” compares businesses within one country, then takes the weighted average; the 95% confidence intervals in that column run from -2.6 to +2.5 points, and every one of them contains zero.
CheckPassNamedFailNamedGap, same country
All three checks3062.3%1,1642.2%-0.6 points
City in the title5053.0%9651.9%+0.4 points
Structured data block7422.0%7332.5%-0.7 points
Readable address on the page7793.1%6961.3%+0.3 points

None of these differences can be told apart from zero. That does not prove the fixes are useless: with 33 named businesses, an effect of one or two points would go unnoticed. It does prove they are not what separates the named businesses from the rest.

In the UK, where the names come out

The UK is the only market large enough to look at closely: 336 businesses, 16 named. The finding is the same there, sometimes reversed. Sites passing all three checks are named 4.0% of the time, the others 5.1%. Those publishing a structured data block: 3.8%, against 6.0% without.

British vets make it concrete. Out of 71 practices, 7 are named. Three of those seven write their city in their page title. Among the 64 the assistant ignores, 23 do. The fix we recommended yesterday does not tell one group from the other.

And the 18 sites in the sample that shut the door on answer crawlers? One of them is named anyway. That is too few to conclude anything except this: the assistant does not need to read your site to cite you. It can know you through what others write about you.

What really separates them: the trade, then the country

Sorted by trade, the same 1,475 businesses no longer differ by tenths of a point.

Share of businesses named, for every trade with at least 40 audited businesses. Same sample, same dates. 87 of the 89 dentists in the second row are British.
TradeMarketsAuditedNamedShare
VetUK, Singapore8589.4%
DentistUK8944.5%
LawyerUK, Singapore5523.6%
ArchitectUK, Singapore4012.5%
DentistFrance, Switzerland, Belgium4112.4%
AccountantUK, Singapore5112.0%
Estate agentUK, Singapore4200%
Real estate agencyUAE, Saudi Arabia, Qatar6700%

A vet has close to a one-in-ten chance of being named, an estate agent none, and their page title has nothing to do with it. Then the country: 4.8% in the UK, 4.7% in Switzerland, 2.7% in Singapore, 1.6% in France, and zero out of 220 businesses in Japan, Saudi Arabia and Qatar.

Neither the trade nor the country is a choice. But both tell you where to look. Our earlier measurements showed it: for a given trade, assistants lean on the same directories and the same review sites from one city to the next. That is where the list is decided, not on your home page.

What this measurement does not prove

That the three fixes are useless.

A page a machine cannot read cannot serve as a source, and the three fixes take thirty seconds. They are still worth doing; they are not enough.

That there is no effect at all.

33 named businesses is few: an effect of one or two points would not show. The intervals say so.

A cause.

We are looking at two measurements side by side, not an experiment. A site that changes its title tomorrow is not a business we observed.

A general law.

The sample is our prospecting file: small businesses that publish their email address on their website, in twelve countries. Large brands are absent, and the US counts a single business here.

What we are changing in our own advice

Yesterday we wrote that three free fixes make a website readable. That is true, and readable does not mean named. We keep the three fixes, filed under hygiene, and add the steps that matter more, also free:

Ask your customers' question yourself.

In a private browser window, ask ChatGPT “who is the best [your trade] in [your city]?”. Three times, a few hours apart: a single answer measures nothing.

Write down who is named, and where the assistant got them.

The sources cited under the answer are, for your trade, the doors into the list.

Be where your named competitors are

: the same directories, the same review sites, with your city written out in full.

All of this can be done without us. If you would rather have the measurement done properly, across several assistants and several runs, that is what our free audit does.

Frequently asked questions

Should I still put my city in my page title?
Yes, because it is free and a machine has to be able to place you before it can cite you. But across 1,475 businesses, writing the city in the title does not measurably change the chance of being named by an AI assistant once the country is accounted for: +0.4 points, with an interval from -1.4 to +2.2.
What makes an AI assistant name one business rather than another?
In our data, the trade first: 9.4% of vets are named, no estate agent is. Then the country: 4.8% in the UK, zero in Japan and the Gulf. Our earlier measurements show that, for a given trade, assistants rely on the same directories and review sites from one city to the next.
Can a site that blocks AI crawlers still be cited?
Yes. In this sample, 18 sites shut the door on answer crawlers and one of them is named. The assistant can know you through third-party sources without ever reading your site. That is too few cases to make it a rule.
Do these results contradict your 10 September article?
They complete it. On 10 September we measured what a website tells a machine; here we check whether that predicts being named. Within the same country, it does not. The three fixes remain useful and free, but they are not what decides the list.