The gap no Western tracker covers
Your brand in Alisa AI, GigaChat and DeepSeek
Every AI visibility tool on the market watches ChatGPT, Gemini, Claude and Perplexity. If your customers are in Russia or the CIS, that list misses the two assistants they open every day — and you are measuring a market you are not selling to.
Why this is a real blind spot
AI assistants are not interchangeable. Each one was trained on different data, retrieves from different sources and is embedded in a different product surface. A brand can be named confidently by ChatGPT and be completely unknown to a local assistant — and the reverse happens just as often.
For a company selling into the Russian-speaking market that difference is not academic. The assistant a buyer reaches for is the one built into the search engine, the bank app or the smart speaker they already have.
Alisa AI (Yandex)
Alisa is Yandex’s assistant: it sits inside Yandex Search, the Yandex browser, the mobile app and Yandex smart speakers. When someone types a “who should I choose” style question into Yandex, a generated answer with company names can appear above the usual results — and a large share of those queries end right there.
What it means for you: if Alisa does not name your company, a buyer who never scrolls past the answer never learns you exist. No Western visibility tracker will tell you that, because none of them ask Alisa.
GigaChat (Sber)
GigaChat is Sber’s model, reachable through Sber’s own apps and through its API. It is the assistant most likely to be embedded into Russian corporate products, which makes it disproportionately important for B2B: the person asking it is often asking on behalf of a company that is about to buy something.
What it means for you: B2B visibility in GigaChat is a separate result from B2C visibility in Alisa, and it is worth measuring separately.
DeepSeek
DeepSeek is widely used as a cheap, capable general model — both directly and inside third-party products that never mention which model they run on. That last part is why it belongs in a visibility check: your brand can be described to a customer by DeepSeek without either of you knowing it.
ChatGPT
ChatGPT is in the check as the baseline. It is the model everyone compares against, and it is often the one that behaves most differently from the local assistants — which is precisely what makes the comparison useful.
How the measurement works
- Same questions, every model. 16–18 prompts built around your category, asked identically everywhere — otherwise the models cannot be compared.
- 3 repeats per commercial prompt. Generative answers are unstable; one run is an anecdote.
- Counting mentions, not positions. Was the brand named, in what order, next to which competitors, and was the site linked.
- Every raw answer is kept and shown in the report, so any number can be traced back to the sentence it came from.
Model availability depends on API access: the report always names the models that answered, so a model that was not connected can never be silently counted as “your brand is missing”.
What to do with a bad result
Local models lean heavily on locally available sources. In practice, becoming visible to them looks less like classic SEO and more like being properly described in the places they can read: your own site written in plain, factual language; consistent naming and category wording; presence in local catalogues, directories and reviews; content that answers the comparison questions buyers actually ask.
That work is what our agency does after the measurement — GEO promotion in AI answers. The measurement itself stays free and independent of it.