Yandex says Alice AI quick answers on Search hit 49.5 million monthly users in Q2 2026, covering 42% of queries — up seven percentage points quarter on quarter. That is not a side experiment anymore; it is a primary results experience for a large slice of Russian-language search. Google’s AI Overviews debates in English Twitter are useful theatre; this is a live, measured adoption curve on a major engine.
The company credits compact Alice AI Search models (MoE + encoder-decoder) built for Search latency. Blind pairwise tests allegedly favour the new stack over Google AI Overviews 56.7% of the time, with stronger results on long-context, medical, and legal queries. Links to sources stay in the answers, and Yandex claims clarifying follow-up queries roughly doubled over six months — which, if true, means generative answers are expanding sessions rather than only truncating clicks. That last claim is the one CIS SEOs should pressure-test with their own analytics: more follow-ups can mean more branded exploration or more zero-click resolution depending on vertical.
For Yandex SEO, the practical read is familiar: when nearly half of queries can resolve in an AI block, classic blue-link CTR is no longer the whole game. Structured, citable pages that survive extraction matter. So does monitoring when Alice answers your topic without sending traffic — and when they do. Compare to Naver’s AI Tab or Baidu’s AI citation reporting: each market is building first-party instrumentation around generative surfaces at different speeds. Yandex is ahead on consumer penetration numbers; Baidu is ahead on giving webmasters a dashboard label for AI citations.
Pair this with Yandex’s earlier AI-Blender work on SERP block layout and you get a search UI that decides format and ranking of modules in one pass. International teams ignoring Yandex because “it is not Google” are leaving a large AI-SERP laboratory unread.
Caveats: vendor win rates versus Google are marketing until you reproduce them; medical/legal strength claims raise hallucination and liability questions that classic SEO reports never covered; and 42% coverage does not mean 42% of your commercial queries. Log Alice presence on money keywords weekly, keep source-linked pages easy to quote, and report “answered without click” as its own KPI beside organic sessions.
Reporting hygiene matters here. If Alice covers 42% of queries globally inside Yandex, your brand’s coverage rate on commercial head terms may be far higher or far lower. Build a sample of 100 money queries, note Alice presence, citation link, and whether a click still happens. Repeat monthly. That little panel will beat any vendor win-rate chart in a client meeting.
Also separate brand and non-brand. Generative answers that resolve non-brand research can still leave brand navigational queries relatively clicky — or not, if the dialog tab absorbs follow-ups. The 2× clarifying-query claim is optimistic for engagers and worrying for publishers. Measure both cohorts before you rewrite the CIS traffic forecast.
When you cite the 49.5 million figure externally, always pair it with your own sample-based coverage rate so clients do not import Yandex’s average as their forecast.