Search with Yandex AI

Yandex AI produces a written answer to a query, automatically or when the user asks for it. It breaks a complex question down, finds sources, analyses them in real time with a generative model, and combines the result into one answer with references to the original sources — the only exception being solutions to school maths problems. A follow-up search box sits beneath the answer.

How pages get used

Answers are generated from pages Yandex has already indexed. Yandex states what makes inclusion more likely in one sentence: information that is well structured, informative and grammatically correct.

Nothing else is on offer — no markup, no submission, no opt-in. Being usable by a summariser is the same work as being readable by a person: state the fact plainly, keep the structure clean, do not bury the answer.

The opt-out, which is the concrete part

To keep a site’s content out of AI answers, disallow YandexAdditional oder YandexAdditionalBot in robots.txt:

User-agent: YandexAdditionalBot
Disallow: /cats/mycat.html

The change takes effect when the robots recrawl the page — typically 2 to 14 days. Note that this is page-level as well as site-level: a single URL can be excluded while the rest remains eligible.

The decision behind the directive

Opting out removes the citation as well as the summary. For a publisher whose value is the full article, that may be right; for a business whose pages answer practical questions, being the cited source is usually worth more than the click it might have replaced. It is a commercial decision, and it is reversible — but slowly, on the same 2-to-14-day cycle.

What to measure

Impressions can rise while clicks fall when an answer resolves the query on the results page. Reporting the two separately keeps that from being read as a ranking loss, because the position has not moved.