Skip to content
Waseit
← All articles
AEOGEO

The door is open. Nobody says where.

3,809 business websites read across twelve countries: only 37 shut the door on answer engines. The gap is elsewhere. Eight home pages in ten say who you are, three in ten say where.

1.0%

of sites block answer engines

September 10, 20269 min read
In this article
  1. What we measured, and how to repeat it without us
  2. The door is open, and that is not good news
  3. What is missing is the place
  4. Twelve countries, and they do not agree
  5. What our own tool taught us, and what we nearly published wrong
  6. Three moves, free, in this order

Short answer: the fear being sold to you is the fear of blocking, and it is wrong. Across 3,809 business websites read in twelve countries, 37 shut the door on answer engines. Thirty-seven. The real gap is duller: 85.4% of home pages say who you are, and only 30.2% say where. An answer engine answers a local question. With no place, it has nothing to answer.

30.2%

name their city in the page title

15.7%

pass all three checks

Unblocking a crawler you never blocked changes nothing. Writing your city into your page title takes thirty seconds.

What we measured, and how to repeat it without us

On 9 and 10 September 2026 we read 3,809 business websites across twelve countries: practices, clinics, shops, agencies. No engine queried, no API key, not one cent. Two to three public HTTP requests per site, the kind anyone can make from a browser.

Three checks, and one principle: measure only what the business itself can verify in thirty seconds.

Are answer engines allowed to read the page?

Does `robots.txt` permit OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot and Perplexity-User.

Can a machine tell who you are and where?

A `JSON-LD` block of type `LocalBusiness` or `Organization`, carrying name, address, phone, opening hours.

Does the page title name the business and the city?

The first line an engine reads, the only one it always shows, and the one your visitor sees in their browser tab.

The door is open, and that is not good news

Two families of crawlers must be kept apart, and confusing them turns a measurement into a scarecrow. Answer crawlers fetch the page at the moment someone asks the question: blocking them removes you from the answer. Training crawlers feed the models: blocking them is a perfectly legitimate business decision and costs no citations.

37 sites out of 3,809 block answer crawlers, that is 1.0%. A further 111 block training, 2.9%, which we do not count as a fault. Put plainly: in 99 cases out of 100, the door the market offers to unlock for you was never shut.

Training crawlers blocked, across 3,809 sites read in twelve countries. Blocking training is a legitimate choice: it costs no citations.
CrawlerSites blocking itShare
GPTBot992.6%
CCBot922.4%
ClaudeBot872.3%
Google-Extended802.1%
Applebot-Extended772.0%

What is missing is the place

85.4% of home pages name the business in their title. 30.2% name the city. That fifty-five-point gap is the whole subject of this article.

An answer engine almost never answers “what is this company”. It answers “a dentist in Rouen”, “a letting agency in Kraków”, “an accountant in Geneva”. To place you inside that answer it has to bind two things: a name and a place. Naming yourself is the floor, and eight in ten do it. Placing yourself is what almost nobody does, and it is exactly what the question asks for.

The structured block is present on 45.9% of sites. Better than the city in the title, and still fewer than one site in two. In total, 15.7% pass all three checks; 41.6% pass only one.

Twelve countries, and they do not agree

The spread between markets is the sharpest result of the measurement, and it follows neither wealth nor digital maturity. It follows a writing habit.

What an answer engine can read, by market. Eleven markets of at least 40 sites, read on 9 and 10 September 2026. The remaining 35 sites, 33 of them American, are outside the table for want of numbers.
MarketSites readThree checks out of threeMachine-readable identityCity in the titleAnswer engines blocked
United Arab Emirates23932.2%58.6%43.5%0.0%
United Kingdom40228.4%52.7%48.5%1.7%
Singapore34623.4%45.7%39.0%0.9%
Poland48718.1%48.7%33.7%0.8%
Switzerland49517.6%46.5%34.5%0.8%
France29017.2%54.8%30.0%0.3%
Brazil3468.7%36.7%17.6%1.2%
Belgium3828.6%38.5%23.8%1.8%
Qatar2766.2%38.8%12.3%0.7%
Saudi Arabia2912.4%41.2%4.1%0.7%
Japan2200.9%41.8%1.8%0.0%

Japan is the extreme case: 4 sites out of 220 name their city in the title. And yet 41.8% publish a structured block there, a rate comparable to Belgium or Qatar. The technique is present; the sentence is missing. The same holds for Saudi Arabia, at 4.1%.

At the other end, the Emirates and the United Kingdom lead the table for no mysterious reason: their page titles already read like a listing - business, trade, city. That is not a technological lead, it is a writing convention.

What our own tool taught us, and what we nearly published wrong

A first version of this measurement put Japan at 0.9% of titles naming the city, and Saudi Arabia at 3.8%. Those figures were wrong, and the fault was ours.

Our title reader split text into words on non-Latin characters. On “京都デンタルクリニック” it found no words at all, so no comparison was possible, so it answered “the title does not name the city”. That was not an observation about those pages: it was a deduction from our own inability to read them, delivered with the confidence of a measurement. Japanese, Arabic, Korean, Chinese, Russian, Greek and Thai all fell into the same hole.

We fixed it - these scripts are written without spaces between words, and the comparison that makes sense for them is containment, not token equality - then re-read the 346 sites concerned. The result confirmed the finding rather than erasing it: Japan moves from 0.9% to 1.8%, Saudi Arabia from 3.8% to 4.1%. The gap was real; it was merely exaggerated by our tool.

We added a third state to that check. When a business name is stored in Japanese in our data and the page title is in romaji, the two do not compare: answering “no” would accuse, answering “yes” would flatter. The check now returns not measurable, and the line disappears from the report rather than inventing a verdict. That covers 25 sites, 0.7%.

A measurement that accuses wrongly is worse than no measurement: it does not cost you the line that is wrong, it costs you the lines that were right.

Three moves, free, in this order

The order matters: the first move settles the other two, and the third is pointless if the first has not been done.

Check your `robots.txt` once, then forget it.

Open `yoursite.com/robots.txt`. If you do not see `Disallow: /` under a `User-agent: *`, the door is open and there is nothing to do. That is 99 sites in 100, and it is the move the market bills you for.

Write your city into your home page title.

“Dupont Practice” becomes “Dupont Practice, dentist in Rouen”. It is the first line an engine reads, the only one it always shows, and seven businesses in ten do not use it. Thirty seconds.

Publish a `LocalBusiness` structured block.

Name, postal address, phone, opening hours. Fewer than one site in two has one. It is the layer an engine reads to know who you are and where, without having to guess.

None of these three moves is for sale. All three can be verified from a browser, on your own site, without us.

Frequently asked questions

Do I need to unblock GPTBot to appear in ChatGPT?
No, and this is the most common confusion. GPTBot feeds model training; the crawler that fetches your page at the moment someone asks a question is called OAI-SearchBot or ChatGPT-User. Blocking training costs no citations. In our measurement, 2.9% of sites block training and only 1.0% block answer engines.
How many sites actually block AI answer engines?
37 out of 3,809, that is 1.0%, in our twelve-country reading of 9 and 10 September 2026. Blocking is therefore a real problem for one business in a hundred, and a sales pitch for the other ninety-nine.
Why does the city in the title matter so much?
Because an answer engine almost always answers a local question: “a dentist in Rouen”, not “who is this practice”. To place you in that answer it must bind a name to a place. 85.4% of pages name the business, 30.2% name the city: the second figure is the one that decides.
Do these figures apply to my country?
The spread between markets is wide and worth looking at: from the United Arab Emirates (32.2% clean sweeps) to Japan (0.9%), the ratio is one to thirty-five. The cause is not technical - Japan publishes as many structured blocks as Belgium - but comes down to a convention in how page titles are written.
Can I run this measurement myself?
Yes, entirely, and without us. Open yoursite.com/robots.txt, look at the title in your browser tab, and search for “application/ld+json” in the source of your home page. The three checks in this article require nothing else.