Here is the Yandex-side conversation worth tracking this week — product posts plus Russian webmaster discourse, not Google Twitter.
Product
- Alice AI’s Expert mode and automatic model routing (15–16 September) is the fresh architecture story. Yandex says ~80% of requests stay on the Alice AI LLM; complex multi-constraint tasks escalate. The harness write-up makes the SEO point explicit: a search tool inside the agent loop feeds answers.
- Measurement still centres on Webmaster’s Alice AI visibility / Share of Voice report — last three months, weekly updates, presence-focused (no click/impression console), limited to queries where you already rank high enough in classic Search. Alice answers rebuild from the current SERP, so sources for the same query can flip.
Forums
On SearchEngines.guru, practitioners are still chewing a June experiment that compared Google/Bing top-10 URLs with ChatGPT/Gemini/DeepSeek-with-web-search citations on four English B2B queries (10 runs each). Only 4 of 120 SEO-top positions appeared in neural citations (3.3%), all on the Bing side; Google overlap was zero in that sample; ChatGPT overlapped neither engine’s tops. DeepSeek citations skewed toward press releases and news wires. Replies immediately asked for larger N and for comparable Russian measurements against Alice — which is the right instinct if you live in CIS SERPs.
What to do
Log Expert-mode-prone commercial queries (travel, shopping, finance) weekly; watch Alice SoV bands, not vanity position screenshots; and stop assuming classic top-10 URLs automatically become answer citations. If your monitoring stack only tracks Google AI Overviews, you are blind to the engine where Alice already covers a huge share of sessions.
This recap deliberately skips Wordstat Cloud API chatter already covered earlier this week. The live debate is about answer presence and citation overlap, not keyword-tool wrappers. Pair the forum anxiety with Yandex’s public Expert-mode docs and you get a coherent week: more agentic answers upstream, still-weak public proof that blue-link winners become citations downstream.
Also worth bookmarking for next week’s crawl: whether Alice SoV moves after Expert mode’s wider routing, and whether SearchEngines.guru posters publish a Russian remake of the 3.3% overlap study. Until then, treat English LLM citation studies as hypotheses for Alice — not answers.
If you only have bandwidth for one operational habit this week: open Webmaster Alice visibility on Monday, screenshot SoV bands for your top money themes, and compare them after Expert-mode routing has had a full week in the wild. That beats another Slack debate about whether 3.3% English overlap “means Alice ignores SEO.”