Advertising in AI Assistants: How Yandex Monetizes Alice and What It Means for Brands

Yandex has published monetization figures for its AI assistant Alice for the first time: since June 2025, advertising formats in the conversational interface have been tested by over 460,000 businesses, and revenue from all monetization experiments throughout the year exceeded 2 bln ₽. Ads are embedded in text format or as product cards — algorithms analyze not only the user's query but also the entire context of the conversation with AI, which allows showing ads at early stages of the funnel when the person hasn't yet formulated a specific need.

How dialogue context targeting works

The key difference between advertising in AI chats and classic search promotion is accounting for the entire dialogue rather than a single query. Direkt algorithms match the neural network's response with the promoted product or service: if a user asks Alice about weekend plans, the system can embed a café or theater ticket offer, even if the person didn't directly request these categories. According to Yandex data, in 65% of cases this approach delivers ads more relevant to the overall meaning of the conversation than exact keyword matching.

About 10% of all queries to Alice are already related to product selection and purchases directly in the chat — in these scenarios, advertising is perceived as a natural part of the interface. Ads are marked with a special icon and visually separated from organic neural network responses, while not affecting the speed or quality of content generation. The vendor emphasizes: smart speakers and child accounts in the app remain outside the testing of promotional formats.

Economics of AI assistant promotion

The cost of customer acquisition through Alice's chat surfaces is comparable to campaigns in classic Yandex search — pricing is determined by the same VCG auction. The format shows the highest conversion in retail, HoReCa, and tourism: these are areas where users most often transition from a search query to dialogue-based option selection. Advertisers receive access only to anonymous aggregated data — number of impressions, clicks, and conversions — without personal user information.

To launch promotion in Alice, it's enough to create a search campaign in Direkt: algorithms automatically select ads relevant to the query and dialogue context. No separate configuration or budget increase is required — the system itself distributes impressions between classic search and AI surfaces. This lowers the entry barrier for small and medium businesses already working with search promotion.

Advertiser budgets follow audience attention, which increasingly spends time in AI services — deep understanding of interaction context allows creating a trusted, relevant advertising touchpoint.

Global market context: who else is monetizing AI

Yandex is among the world's pioneers in AI assistant advertising: testing on the Russian market began in 2025, while OpenAI launched ads in ChatGPT only in early 2026 and has already opened a separate advertising cabinet. Google is integrating ads into AI Mode and AI Overviews of its search, hinting at expanding this model to its Gemini neural network. Microsoft has been testing ads in Bing Chat since 2023; today the format works in Copilot as well.

Training a foundational language model costs vendors tens of millions of dollars — expenses for datasets, infrastructure, and engineering teams require quick payback. Yandex is simultaneously testing three monetization models: advertising, subscription, and transactional. Combined revenue from experiments throughout 2025 exceeded 2 bln ₽, but the company doesn't disclose the share of each direction in this figure.

How users view ads in neural networks

According to OMI research (April 2026), 70% of Russians will accept ads in AI provided they're relevant to their queries, 57% — if they're organically integrated into the interface. 62% of respondents choose free neural network formats with ads, and only 10% are willing to pay for a version without promotional integration. These figures confirm the viability of a hybrid model: basic functionality remains accessible to a wide audience through ads, while paid subscription provides expanded capabilities and no ads.

The key acceptance factor is targeting quality. If an ad solves the user's problem or offers a relevant alternative, it isn't perceived as an obstacle. If algorithms miss the context, irritation grows faster than in classic search: the dialogue interface creates an illusion of personal communication, and a non-relevant ad breaks this feeling more strongly than a banner on a results page.

Marketer's checklist: how to prepare a campaign for promotion in AI chats

• Check the structure of your search campaigns in Direkt: ads should answer a broad spectrum of queries, not just transactional ones — AI context captures early funnel stages.
• Review your semantics: add informational and navigational queries related to your category — the algorithm will match them with the dialogue even if the user didn't mention your brand.
• Prepare product cards with clear titles and descriptions: in a chat, users evaluate an offer in 2-3 seconds; long texts don't work.
• Set up conversion tracking from new traffic sources: Direkt provides aggregated data, but for effectiveness analysis you need cross-channel analytics on your side.
• Monitor CPO and compare it with classic search: if the cost of acquisition is higher with comparable lead quality, adjust your bids or exclude impressions in AI through campaign settings (if such an option appears).
• Test hypotheses with small budgets: start with one product category or service, collect statistics over 2-4 weeks, scale successful combinations.

Risks and limitations of the new surface

The main problem with AI chat advertising is algorithm transparency. Yandex doesn't disclose exactly how the system decides to show an ad in a specific dialogue: an advertiser sees only final metrics but can't trace the chain "query → context → impression." This complicates campaign optimization and increases dependence on algorithm black box.

The second difficulty is attribution measurement. If a user saw an ad in Alice but made a purchase days later through another channel, standard attribution models won't account for the AI surface's contribution. Advertisers will have to invest in multi-channel analytics or rely on vendor aggregated data, which doesn't always reflect the full customer journey picture.

The third limitation concerns ad labeling: all ads must comply with advertising law requirements, including mandatory marking. Yandex claims that promotional materials are marked with a special icon, but the mechanics of submitting information to the Federal Internet Advertising Registry for dialogue formats hasn't been publicly detailed yet — it's important for brands to ensure everything is technically implemented on the platform side.

Frequently asked questions

How to launch ads in Alice AI

Create a search campaign in Yandex Direkt — algorithms will automatically select relevant ads for display in Alice's chat. No separate configuration or additional budget is required: the system itself distributes impressions between classic search and AI surfaces via VCG auction.

How much does advertising in Yandex AI assistants cost

The cost of customer acquisition through Alice is comparable to campaigns in classic Yandex search and is determined by the same auction. The specific price depends on competition in your niche and ad quality — pricing principles don't differ from search promotion.

What user data is available to advertisers in Alice

Advertisers receive only anonymous aggregated data: number of ad impressions, clicks, and conversions. Personal user information, dialogue history, and detailed impression context are unavailable — the vendor restricts data access at the platform level.