AI characters and virtual brand avatars

Is an avatar needed for a brand at all

Most articles about AI characters begin with excitement about the technologies — motion capture, LoRA models, speech generation. But the question of 'how to make' is secondary to the question of 'why'. An avatar solves specific business tasks: scaling content without increasing the budget for filming, reducing reputational risks (a virtual character will not have a scandal in personal life), omnichannel — one image works in advertising, in an app, and in a stream at the same time.

Before launching a project, it is worth going through several forks:

1. Check whether the avatar solves the task cheaper or better than a live person — a blogger, actor, or staff member in the frame.
2. Evaluate the content frequency: if 1-2 videos per month are needed, the return on investment for creating a character is questionable; with a stream production (as with Rafinad, where speed increased 3–5 times), the economy begins to work.
3. Determine the audience segment and select the level of human-likeness to match their expectations — more on this below.
4. Decide in advance about revealing the artificiality of the character — this is not a technical detail, but part of positioning.

The trust barrier: why two-thirds of the audience are skeptical

Distrust of AI characters is rarely related to the fact of their artificiality. It arises from the feeling of substitution — when a brand passes off synthetic content as organic, human. Research shows: 57% of consumers value content quality more than its source. That is, the audience is ready to forgive a brand for using an avatar if the result works — entertains, informs, and solves the task.

The barrier is overcome not by justifications, but by transparency from the start. Aviasails does not hide that Aviaslav is a character; on the contrary, the company uses three different personas for different segments (for women, for IT professionals, for youth), and this works: Ad Recall +7%, brand recognition — 75%. Openness becomes part of the brand's character, not a weakness to be hidden.

What strengthens trust, and what destroys it

• Works: clear designation of the character as a mascot or AI hero from the very launch.
• Works: consistency of the image — if the character is honest about its nature once, it should not pretend to be alive in other channels.
• Destroys trust: post-factum exposure, as in The Guardian's investigations into hidden virtual influencers — in this case, the reputational blow is stronger than the news of AI usage itself.
• Destroys trust: mismatch of the character's tone with the context — humor from a cat-mascot in a serious financial consultation causes rejection.

The Uncanny Valley: when human-likeness works against the brand

The effect of the Uncanny Valley is not a myth from the video game industry, but a real factor of AI character effectiveness. The higher the degree of anthropomorphism of the avatar, the higher the audience's demands for its authenticity. A semi-schematic cat or cartoon mascot does not cause dissonance — the brain immediately recognizes the conventionality of the image and does not attempt to match it with a real person. A photorealistic digital human with even minor artifacts in mimicry or voice provokes the feeling of 'something is wrong,' and this feeling is transferred to the brand.

This leads to a strategic rule: high anthropomorphism requires a trust evidence base — expert credentials, transparent disclosure of technology, quality implementation without visual bugs. Low anthropomorphism, on the contrary, plays on novelty and emotional frames — a mascot will forgive conventionality if it is funny, charming, or useful. SberKot and Athena win precisely due to moderate stylization: they do not claim photorealism and therefore do not fall into the valley.

Where AI characters work and where they fail

The effectiveness of an avatar strongly depends on the type of content. Product demonstrations, explanations of tariffs, answers to frequently asked questions — this is the area where an AI character is perceived neutrally or even positively, because the audience is not looking for emotional authenticity here. A completely different story is lifestyle and motivational content, where the value of the message is built on personal experience.

The conclusion is simple: an avatar is strong where the audience expects functionality and systemacity, and weak where it expects genuine personal experience. SberKot, in this sense, is a model of precise positioning: it does not claim to be a friend or advisor on personal finances, but rather a mascot-guide through the product, which allows lowering the barrier to entry for new users.

Real-time management of an AI character

VTubing and live avatar technologies transform an AI character from static content into an interactive format. The principle of operation: an actor or operator transmits movements and facial expressions to a digital model via motion capture, and the speech processing system (AI speech processing, TTS) synthesizes the character's voice in real time — either by voice-over of a live person with the desired tone, or by directly generating responses.

Russia's first VTubing stream with the character E.M. Vi gathered 8 million viewers — the avatar answered audience questions live using a motion capture and AI speech processing setup. This creates formats unavailable to regular content: AMA sessions on behalf of a brand without risk to the speaker's reputation, simultaneous presence of the 'character' across multiple platforms, integration into metaverses and gaming platforms without physical constraints. Technically, the entry barrier is reduced — what previously required weeks of post-production, Rafinad-type studios now accomplish in one day, explaining the multiple increase in content release speed among major brands.

To reveal artificiality or to mask it

This is the only question in AI character strategy where there is no universal answer, but there is a clear regulatory trend. As of August 2026 year, the AI Act comes into force in the European Union, requiring AI-generated content to be labeled. Russia currently lacks a direct analog of this law, but platforms are explicitly implementing requirements — YouTube already mandates a separate flag for synthetic content.

According to industry experts (particularly Clarissa Mansbridge), 40–60% of content from major brands can now be created with AI involvement — and a significant portion of this volume is hidden out of fear of losing audience trust. This is a strategic mistake. The risk of exposure and subsequent investigation (following the example of The Guardian's materials on unscrupulous virtual influencers) costs reputation more than a honest disclosure of the character's nature at the project's outset. Revealing artificiality, built into the brand story from the beginning, transforms from a threat into an individuality element — as Avisail and Sber have done, and as more and more brands will have to do as regulation tightens and labeling requirements increase.