Home

Product concept · 2026

Kiki

A Kinopoisk AI assistant that finds films for even the most nuanced request.

Viewer Context

Problem

Choosing a film often begins with a feeling: you want something cozy, like in childhood, a story with an unexpected ending, or a movie that will remind you of your favorite film. Such desires are difficult to formulate in the search bar and almost impossible to convey using standard filters.

As a result, the viewer spends his free time on an endless search. Ratings, genres and viewing history do not always reflect what you want right now. And when a suitable film is finally found, it may not be included in the Kinopoisk subscription. The search drags on, interest fades, and the user leaves.

Solution

Kiki understands what kind of movie you want to watch, even when the viewer himself cannot really explain it.

It perceives emotions, associations and disparate wishes, turning them into understandable search tags. Together with the usual filters, they help to more accurately convey the mood and find films that really suit the request.

And the “Why Kiki Recommends” section on the movie page explains the recommendation through specific details: atmosphere, plot, characters and experiences that the viewer is looking for.

Business context

Problem

Each unsuccessful search reinforces the user’s feeling that it is difficult to find a movie for themselves in Kinopoisk. Even if a suitable film is in the catalogue, the viewer may not get to it and go to competitors.

For a business, this means lost attention, decreased engagement, and missed opportunities to convert a viewer into a subscriber. At the same time, the service receives less information about the user’s real interests: what attracts him, what emotions he is looking for and why he chooses a particular movie.

Solution

Kiki helps Kinopoisk find a common language with the viewer and reveal the value of its own catalogue.

Using artificial intelligence, Kiki interprets natural language queries and matches them with an additional semantic layer of tags that describe the atmosphere, emotions and impressions of films.

For example, “Harry Potter” is not only “fantasy”, “magic” and “adventure”, but also “strong friendship”, “the fight between good and evil”, “New Year’s atmosphere”, “warmth and comfort”. Such connections allow you to find films based on associations and emotions that a typical genre search misses.

This helps users discover more relevant content within Kinopoisk, and helps the service personalize content more deeply. Today, thanks to AI, there is more and more content competing for user attention, so it is important to better understand preferences and provide everyone with a unique emotional experience.

If the movie is not in the catalog, Kiki suggests “Find in Yandex.” This will allow the user to remain inside the ecosystem and continue to observe his actions, knowing the context of the request.

Hypotheses

Natural-language search shortens the path to a choice

When viewers can describe what they want in their own words, they can choose a film and start watching faster than with filters alone.

Explanation increases trust

If you show which features of the film coincide with your request and taste, then the person will use the function more often and return to the service.

Opportunity changes habits

If you liked the film, but it is not in Kinopoisk, then using the available button the user will more often search for it in Yandex Browser.

How to test an idea

Qualitative

Movie-selection tasks

Ask 5–7 participants to find a film for a specific evening using both the current search and the Kiki prototype. Observe how they phrase the request, interpret the results, and understand the recommendation rationale.

Quantitative

Controlled experiment

Compare Kiki with the current search on comparable traffic. Measure play conversion, time to choice, and recommendation relevance.

What is considered success?

Conversion from search to movie launch

The share of Kiki sessions in which the user started watching in Kinopoisk or searched for a movie and went to the browser.

Median Time to Play

Median time from the first request to the launch of the movie.

Save Rate

Proportion of sessions with at least one movie saved.

D7 Repeat Use

Percentage of viewers who returned to Kiki within seven days.

Recommendation Relevance Score

Assessing how well films suit the viewer's needs.

Expected effect

Conversion from search to browsing

Personalized selection will make it easier to find interesting movies.

Regular user growth

The ability to select films in this way will allow you to retain and gain new audiences.

Improving personalization algorithms

An increase in metrics will show the usefulness of the function and open up opportunities for in-depth personalization.

Transitions to Yandex Browser

Films outside the Kinopoisk catalog with a button to go to Yandex Browser will increase the activity of users of the main service and, possibly, attract new ones.