Yandex Alice AI Adds Expert Mode and Automatic Model Routing

Yandex is no longer asking users to pick a “smart” toggle and hope. On 16 September 2026 the company said Alice AI now routes each request to the model it thinks the task needs. Everyday asks stay on fast models. Multi-constraint planning, multi-document work, calculations and step-by-step plans can flip into Expert mode automatically — or you can force it yourself.

The English product note is thin; the useful detail is in Yandex’s own Medium engineering write-up and the Russian Habr mirror. A classifier estimates resource need and picks a model. Yandex says the Alice AI LLM currently handles about 80% of requests. Expert mode’s model set is tested daily and can mix Alice-family models with open-weight models hosted on Yandex servers. That is not “one frontier model with a nicer UI.” It is an ensemble with a traffic cop.

More important for search watchers is the harness story. Yandex separates the weights from the product layer: system prompts, tools (including live search and a Python sandbox), an agentic loop that accumulates tool results, a context manager that forces a final answer before the window fills, file parsers plus a VLM for image/PDF pages, RAG over long documents, plus a summarizer and renderer for the UI. File-heavy prompts default into Expert mode. If you care about Alice answers citing the open web, the search tool inside that harness is the retrieval customer — not the headline parameter count.

Yandex also published its internal “Difficult Benchmark” (v0 ≈ 50 hard, real-shaped tasks). Scores across 50-plus runs ranged from 23 to 71/100; GPT-6 Astra Pro topped the board in their telling. Open-weight GLM 5.2 high scored 48 with Yandex’s harness versus 43 on the developer’s own configuration. Habr’s trip-planning examples (Istanbul for two inside ₽120,000; a robot vacuum under ₽40,000 that handles carpets and cat hair) show the commercial intents they want Expert mode to own.

For Yandex SEO this reinforces a pattern already visible in Alice quick answers and AI-Blender layouts. Visibility is increasingly about being extractable into an agent loop that may search, parse and plan before it ever shows a blue link. Structured facts, clear entities and pages that survive passage extraction beat keyword-stuffed intros. Watch shopping, finance, law, healthcare and travel — the verticals Yandex itself flags as Expert-friendly.

Bottom line: Expert mode is productisation of model routing, not a new search index. Combined with Alice’s mass quick-answer footprint, it is another reason CIS teams should log module layouts and citation presence weekly instead of obsessing over a single organic position screenshot. Pair it with Webmaster’s Alice visibility / Share of Voice report if you need a first-party presence metric while clicks stay opaque.

Dan Taylor is the founder and editor of International Search Insights (ISI), the publication behind ReadIsi, and a Partner and Head of Innovation (Search & AI) at SALT.agency. He has worked in search for more than a decade, in-house and agency-side, and is regarded as one of the West’s leading voices on search engines beyond Google, including Yandex, Baidu, Naver and Cốc Cốc. On ReadIsi, Dan covers the engines and AI answer platforms that English-language coverage tends to miss: Baidu and ERNIE, Yandex and Alice AI, Naver, Cốc Cốc, Shenma, Sogou, WeChat and Weibo search, plus DeepSeek and Alibaba’s Qwen. He also keeps dated records of documented changes, such as the Yandex, Baidu and Naver algorithm update histories and the DeepSeek, ERNIE and Qwen model update histories. His approach is to take learnings from global search engines and apply logic with a high evidence threshold. A claim needs a source, patterns seen in one market are treated as a hypothesis until they are checked in another, and where the evidence is thin he says so. For brands, that means advice built on what can be shown, with measurement that separates being cited from being chosen. Dan is credited with coining the term “edge SEO” and won the inaugural TechSEO Boost research competition in 2018 for his work on CDN and edge changes for SEO. In 2026 he was credited with coining “Infinite Tail” to describe how keyword research is changing with AI search. He is the author of AI Search Marketing For The Travel Sector, writes for Search Engine Journal, Search Engine Land and Travel Marketing Guy, and has advised technology, SaaS, travel and enterprise ecommerce brands, including Cloudflare, GitLab and Airbnb. He has given multiple talks and panels at the International Search Summit in Barcelona, and has also spoken at Optimization Moscow, brightonSEO, Google Search Central Live Deep Dive Asia in Bangkok, Athens SEO and the Asia SEO Virtual Summit 2026, where his talk looked at how Baidu, Naver, DeepSeek, Tencent, Cốc Cốc and Yandex built their own AI answers.