New research from Moscow agency Ashmanov & Partners tracks a fast shift in Yandex’s commercial search. In April 2026, an Alice AI answer showed up on one in a thousand ecommerce queries in its sample. By July, Alice “quick answers” were appearing on about 15% of them, and often above everything else on the page.
The figures come from the agency’s latest report, summarised by CNews on 6 October. The full report, “How Alice AI answers commercial queries”, is billed as part three of its series on Alice and commercial search. We missed it on Tuesday, but it’s the most useful set of numbers on Yandex AI search and ecommerce we’ve seen this autumn, so it’s worth covering now.
Quick answers: cached, informational, and climbing
Three findings stand out:
- They aren’t live. Ashmanov says quick answers on commercial queries are almost always “offline”. Unlike full answers, and unlike AI blocks on many informational queries, they aren’t rebuilt from real-time search each time, and they “may not change for weeks”. Fix a product page today and Alice’s summary in the SERP may still describe the old one next month.
- They’re informational, not shopping. They compete visually with links to product pages, but for now they don’t replace them. Yandex can add store and product sections to full answers in the Alice AI tab, but it is “in no hurry” to put commercial elements into quick answers on the main results page. CNews links that caution to monetisation, which makes sense, because that SERP is where Yandex earns its ad money.
- They’ll keep spreading. The agency expects the share to grow on high-volume product queries where intent isn’t clear, giving “Dyson hair dryers” and “car from China” as examples.
Full answers: tighter sourcing, and product cards that aren’t yours
The more uncomfortable findings are about full answers in the Alice AI tab.
Fewer queries behind each answer. After the June update, Alice generates fewer derived queries per answer: about four per original query in April, and just over two in July. Source selection is tighter too. The report finds five pages per query, more than half of them from the organic top five and over 90% from the top ten. The agency’s report page adds that over the past six months, the share of sites outside Yandex’s top ten appearing in AI answers has “noticeably decreased”. Classic Yandex SEO still decides who gets cited.
Two kinds of commercial section. “Store” sections are built from the commercial sources Alice selected, and sometimes only one or two of them make it in. “Product” sections work differently. Source pages are only a starting point for choosing models and building queries to Yandex Products (Яндекс Товары). The model details and offer cards come from Yandex’s own service, not from your page. In Ashmanov’s words, the store whose page was used as a source gets “no advantage”, and the card can link to a different shop. Product sections became noticeably more common after the big update on 25 June.
So a retailer can do the SEO work, get picked as a source, and then watch Alice send the buyer to a competitor’s offer in Yandex Products. If that sounds familiar, it’s the same pattern we saw when someone scraped 295,000 Neuro answers and found Dzen out-cited Wikipedia. Yandex’s own properties keep winning the space inside its AI answers.
The shopper survey
The agency also surveyed online shoppers. CNews doesn’t give a sample size, so read these as indicative. Shoppers were 30% more likely to notice the AI block than the search results. 69% agreed that AI services speed up finding products, and 74% use AI answers in search engines. Two-thirds had bought something straight after using an AI service (14% often, 52% sometimes). 38% said what they bought often or almost always matched the AI’s recommendation, and another 45% said it sometimes did.
What to do with this
- Don’t neglect informational content on shop sites. Quick answers on commercial queries are informational. Buying guides, comparisons and spec explainers on your category pages are what get cited there.
- Get into Yandex Products. If product sections pull offers from Yandex’s own service, a feed there is the way into the card, whoever’s page got cited.
- Allow for the lag. Cached quick answers mean price or stock changes can take weeks to show. Check the important queries by hand after any big catalogue change.
- Top ten first. With over 90% of sources coming from the organic top ten, the familiar Yandex ranking work still comes first, whatever the AI visibility tools say. Yandex’s own course on getting cited in Alice says much the same.
My view: 15% sounds modest until you remember it was 0.1% in April. With tens of millions of people already using Alice quick answers, Russian ecommerce SEO now has an AI layer that is cached, sourced from the top ten and steered towards Yandex’s own product cards. Plan for that. More in Yandex AI.