AI engines cite you only if you have a page for the exact question

The test

Twelve regions, five AI engines, web retrieval switched on, the same phrasing every time: "top QA companies in" followed by the region. One variable changed from region to region: whether our site had a page for that region.

The result, twice

Six regions had a page and six did not. With a page, we were cited in 20 of 30 engine runs on the first measurement and 25 of 30 on the repeat. Without a page, we were cited in 0 of 30 both times. Every region without a page scored zero on every engine.

Mentions behave differently: counted loosely, we were mentioned 21 times out of 30 with a page and 2 out of 30 without. That is why we count citations only.

What it does not prove

• A page is necessary, not sufficient. One engine cited us in every region with a page. Others missed some, and which regions missed varied between runs.
• The engines are API models. One "engine" is a model reached through a router with retrieval on. It is not a measurement of the consumer chat product.
• The cited pages are list pages. That format earns few clicks in ordinary search, so do not rewrite your pages to chase it.

The conclusion we had to retract

Earlier, our monitor reported that four of the five engines could not cite anyone. That was our own configuration: we had never switched web retrieval on, and the blanks were being read as facts about the engines. The monitor now reports an engine that returned nothing separately from an engine that never names us. An engine that errors on every query is not an engine that never mentions you.

What to do with this

• Write the page for the exact noun in the question. The region, the product, the standard.
• Measure with the same wording and count citations.
• Write down what your probe could not see. Zero and "could not look" must not look alike.