Your avatar is pretending to advise customers. That is the risk

A customer clicks “Get advice” on your website. Suddenly, a friendly face appears. Is it a person? The question may stay unanswered for only a moment. Yet that moment can shape whether the avatar feels helpful or like a trick. Video avatars can handle repeat explanations, even outside your opening hours. They can also cost you trust if their role is unclear. Where is the line, and what will businesses soon need to disclose?
The awkward moment often comes after hello
An avatar greets a visitor by name, smiles and explains the offer. Then the visitor asks about an exception, a delivery problem or a recommendation tailored to them. The avatar repeats a line from its script. Suddenly, the friendliness feels like a disguise.
The risk is less about a face being digital. It is about the impression the avatar creates, and whether its answers live up to it. An avatar explaining a standard feature can be useful. One that gives the impression of offering personal advice, then stalls at the first follow-up question, is more likely to damage trust.
Video and voices can now look and sound convincing, even to someone scrolling quickly. Mainstream products and specialist services have also made it easier to create virtual people. For businesses, the technical question of whether an avatar is possible is quickly answered. The harder question is what people expect when they see its face.
Where an avatar can take work off your hands
The most useful place to start is often with a question your team has already answered three times by Monday. How do I set up access? Which option suits my needs? What happens next? If the answer is consistent, an avatar can offer an initial explanation and guide people through set steps.
That can help with software onboarding. The avatar walks users through the first steps, explains features and answers follow-up questions. A manufacturer can explain the differences between product options on its website, then direct people who need advice to the sales team. Recurring training content can also be delivered in several languages.
For a small business, even a short explainer video about opening hours, processes or a common product question may be useful. A larger company can apply the same idea to customer portals or training. In both cases, it must be clear where the explanation ends and a person takes over. Without that boundary, help can quickly become an obstacle.
The face has to fit the job
An avatar can deliver factually correct lines and still feel wrong. A very human-looking face that responds to every question with the same calm expression can be unsettling in a way that a simple explainer video is not. The more personal and consequential the decision, the more that gap matters.
A digital version of a real person is particularly sensitive. If a trainer is to appear as an avatar in a recurring course, there should be clear agreements about how their voice and appearance can be used. There is a warning in the world of social platforms: Instagram removed an account that used an AI image to pose as a public figure. According to a report, Meta prohibits using someone’s likeness to deceive, whether the content is real or AI-generated.
So do not just check whether the avatar looks approachable. Check whether customers understand its role straight away, whether its tone fits your brand and whether the person shown has agreed to the use. A digital copy should never give the impression that a real person is speaking when they are not.
Labelling takes more than an asterisk
A small note at the bottom of a page will not answer every transparency question. Under the European Commission’s guidelines, people should be explicitly told when they are interacting directly with an AI system. For interactive avatars, that points to a clear notice before the conversation starts, not a mention buried in a privacy policy.
The next key date is 2 August 2026. From that date, Article 50 of the EU AI Act applies. The rules cover transparency duties for certain AI systems, including interactive systems and deepfakes. Realistic AI-generated content may need to be labelled. The EU guidelines also provide for machine-readable markings, so AI-generated or manipulated content can be identified.
These are two separate things: users should know they are speaking to AI, and realistic AI content should be labelled or technically identifiable. According to the guidelines, anyone who does not apply the EU Code of Practice must be able to show that alternative measures are equally appropriate. Which requirements apply to a particular use depends on the company’s role and the content. Check the labelling before launch, not after a complaint.
What this means for your business
The best first step is not an avatar. It is a list of the questions your team answers over and over. Mark which have a reliable standard answer and which call for judgement, discretion or a closer look at the individual case. Only the first group is suited to an automated explanation.
Next, define the role in one sentence: the avatar explains the first steps, shows differences between products or delivers training content. Decide what it must not answer, and how customers can reach a person. A clearly visible handover button is often more convincing than an avatar that acts as if it can handle every query.
Before launch, run a small practical test. Ask staff to try typical and difficult follow-up questions. Check whether the avatar knows its limits, whether the AI notice is easy to see and whether the video label is clear. Then ask people unfamiliar with the offer: did they understand who they were speaking to?
Next Monday, a new customer might work through the initial set-up on their own while your team handles cases that call for experience and tact. Or the avatar might add another hurdle after a few questions. The way you define its task and handover will shape which of those situations you get.
- Collect common questions and separate standard cases from individual ones.
- Set out the avatar’s role, limits and handover to a person in writing.
- Test the AI notice and content labels before publishing.
- Use a real person’s voice and appearance only with clear permission.
Sort out the process before the face
An avatar cannot shortcut its way around unclear processes. If product information is out of date or no one knows who takes over a difficult case, a friendly face only makes the gap more obvious. Someone needs to own the content and conversation, check feedback and keep information up to date.
Klarframe offers video avatars for websites, product advice and onboarding. For businesses considering them, the decision to make first is what the avatar is allowed to explain, and when it hands over to your team. Without that, even the best label is just a sticker on an unclear customer experience.
Frequently asked questions
What is a video avatar?
A video avatar is a digital character shown in video form, using a voice and facial expressions to explain things. Depending on the application, it might greet visitors on a website, explain a product or guide people through onboarding. Interactive avatars can answer follow-up questions. Users should be able to tell that they are speaking to AI, not a person.
Do businesses have to label AI avatars?
The EU guidelines say people should be explicitly informed when they interact directly with an AI system. Article 50 of the EU AI Act applies from 2 August 2026 and sets out further transparency duties, including for certain AI systems and deepfakes. The labelling required in a specific case depends on the company’s role, the content and how it is used.
When does Article 50 of the EU AI Act apply?
According to the European Commission, Article 50 of the EU AI Act applies from 2 August 2026. It covers transparency duties for certain AI systems, including generative and interactive systems, as well as deepfakes. Businesses should check in good time whether they are affected as a provider or deployer, and what notices or markings their specific use requires.
When are video avatars useful in customer service?
They are best suited to recurring explanations with clear answers, such as the first steps in onboarding, product differences or standard training content. A person should be available for complaints, exceptions and personal decisions. A visible handover to staff helps prevent the avatar from promising advice it cannot provide.
Can I use an employee as an AI avatar?
If a real person’s voice or appearance will be used digitally, businesses should agree clearly in advance what that use is permitted for. The avatar must also be recognisable as AI. A digital version should not give the impression that the employee is speaking in person when they are not.
Interested?
Want to know what this means for your company?
Sources
- Klarframe: Video Avatare · klarframe.com
- Nobody is safe from celebrity AI deepfakes · The Verge, 2026-10-15
- AI ‘content creators’ are getting harder to spot · The Verge, 2026-06-07
- Code of Practice on Transparency of AI-generated Content · Comisión Europea, 2026-10-19
- Guidelines on transparency obligations for providers and deployers of certain AI systems · Comisión Europea, 2026-08-06
This article was researched and written with AI assistance and automatically checked against the sources listed before publication. More on our AI transparency
Related solutions
- Video Avatars →
Digital video avatars welcome visitors, explain products and guide customers through content and onboarding.
Klarframe AI Readiness Check
How ready is your company for AI?
About ten minutes, instant result with your stage, a 90-day plan, suitable solutions and notes on the EU AI Act.


