Journal · Method
Creating an AI persona for a brand or a personal brand
An AI persona is a brand character whose face and content are AI-generated. For it to serve the brand instead of damaging it, seven things must be decided before publishing: its role, its identity, its face, its limits, who answers messages, how you disclose it, and who signs off. Here is the method, the risks, and the cases where a real face remains better.
Before you read: Katia Ryze, whose name this site carries, is herself an AI-generated persona. She is not a real person. This article is written by the Ryze team, not by the character.
What is an AI persona, and what is it for?
It is a character that embodies a brand: a face, a tone, a way of speaking. The difference from a drawn mascot is realism. The images and videos look like photographs. That realism is exactly what calls for rules.
An AI persona can serve three purposes:
- Giving a brand a regular face when no spokesperson is available every day.
- Building a personal brand without showing your own face, when the founder prefers to stay in the background.
- Keeping the same tone across several products and several networks.
It replaces neither the product, nor the service, nor the people who answer customers.
The method in seven steps
- Write its role in one sentence. What does the character do for the brand? Present the products, explain a topic, guide visitors. If the sentence does not fit on one line, the project is not ready.
- Write an identity sheet. First name, tone, vocabulary, favourite topics, what she likes and what she refuses. Note on the same sheet that all of this is brand fiction.
- Fix the face. Choose a few reference portraits and list the features that never change. Check that this face belongs to no one (see the risks below).
- List what she says and what she does not say. This is the most useful step. The table below gives a starting point.
- Decide who answers messages. A human from the team, an automated assistant, or both. In every case, the person writing in must know who they are talking to.
- Prepare the disclosure. One sentence for the bio, one for the website, and each platform's label. The details are in our article Openly AI influencer: why and how to disclose it.
- Name a human in charge. Someone reviews before publication and answers if there is a problem. A character cannot carry responsibility.
What an AI persona can say, and what it must not say
| Topic | She can | She must not |
|---|---|---|
| Products | Present, explain, show how it works. | Say "I have used it for years" as a lived testimonial. |
| Results | Quote real company figures, with their date. | Invent revenue, customers or reviews. |
| Personal story | Have a world and a personality, presented as those of a character. | Use an invented story (illness, hardship, success) as a sales argument. |
| Advice | Give general explanations and point to a source. | Give personalised medical, legal or financial advice. |
| Her nature | Say she is AI-generated, without waiting for the question. | Answer "yes, I am real". |
| Advertising | Do a partnership that is flagged as such. | Recommend a product without saying it is paid. |
How do you keep the same face from one image to the next?
This is the most common practical difficulty. A face that changes from one piece of content to the next breaks the effect and looks improvised. Whatever technique is used, the discipline is the same:
- A reference sheet. A few approved portraits, from several angles and in several lights.
- Written constants. Eye colour, hairstyle, distinctive marks, clothing style, usual places.
- A check before publication. Every new image is compared with the sheet. Those that drift from it do not go out.
- Few changes at a time. A new setting or a new outfit, not both with a new framing.
What are the risks?
Misleading the public
This is the main risk. In Switzerland, the Unfair Competition Act treats as unfair and unlawful any conduct or business practice that is misleading (Article 2 of the Swiss Unfair Competition Act). A persona presented as a real customer or a real expert comes dangerously close to that. Only a court can decide a specific case, but the safeguard is simple: say it.
Looking like someone who exists
A generated face can resemble a real person without anyone intending it. The platforms are strict on this point. TikTok does not allow AI content that uses the likeness of a minor, nor that of an adult who is not a public figure without their permission, nor content that falsely shows a public figure in certain contexts, for example endorsing something, even with a label. Before fixing a face, check that it does not reproduce that of a well-known person or of someone close to you, and never start from someone's photo.
Being labelled by someone else
Platforms are getting better at detecting generated content. TikTok may apply an AI-generated content label itself, and Meta adds the "AI info" label when it spots AI image indicators (Meta announcement). On YouTube, creators who repeatedly fail to disclose their AI-generated photorealistic content may be penalised. Better to announce it yourself than to be found out.
Depending on a character
If the whole brand rests on the persona, a change in a platform's rules or a rejection by the public hits everything at once. Keep pages and content that stand without her: products, proof, team.
AI persona or real face: how to choose?
| Situation | Declared AI persona | Real face |
|---|---|---|
| Regular content to present products | Suitable | Suitable, but takes filming time |
| Founder who does not want to be exposed | Suitable, if you say who is behind it | Not wanted |
| Regulated profession or personalised advice | To be avoided | Necessary: a person takes on responsibility |
| Customer testimonial, before and after | To be ruled out | The only honest option |
| Sales meetings, after-sales service | Can point to the right person | Necessary for the relationship |
| Events, press, partners | Limited | Preferable |
The simple rule: the more the customer's decision rests on trust in a specific person, the more an identifiable human is needed. The two can coexist, a persona for content and a human for the relationship.
What we chose for Katia Ryze
We only state here what this site displays. Katia Ryze is the face of the Ryze ecosystem. The home page states that Katia is an AI-generated persona and declared as such, and that the Ryze ecosystem, the software and the results are real and built by the Ryze team. The page header presents her as an "AI-generated persona", and the footer of every page repeats it. An AI persona is not a person and carries no responsibility: real people decide and answer for the company.
What is real is in the portfolio: products that are online, with screenshots. The profile sums it all up. For creating AI characters and content for brands, the ecosystem also has a studio, Ryze AI Studios.
Frequently asked questions
What is an AI persona?
It is a brand character whose face, voice or videos are generated by artificial intelligence. It carries the communication of a company or a founder. It is not a real person, and it must be presented as such.
Can you build a personal brand without showing your face?
Yes, a declared AI persona is one way to do it. It carries the content while the real person stays in the background. You then have to say clearly who is behind it and what is real: the company, the products, the results.
How do you keep the same face from one image to the next?
By fixing a reference sheet before publishing: a few approved portraits, the features that never change, the wardrobe and the usual places. Every new image is compared with that sheet, and those that drift from it are not published.
Can you give your persona the face of an existing person?
No, this should be avoided. TikTok, for example, does not allow AI content that uses the likeness of a minor, nor that of an adult who is not a public figure without their permission, nor content that falsely shows a public figure endorsing something, even with a label. A persona's face must belong to no one.
When is a real face better?
When trust rests on a specific person: regulated professions, personalised advice, testimonials, after-sales service, meetings in person. In those cases, a character cannot take on responsibility in place of a human.
The Ryze ecosystem
Four universes, one way in.
Digitise your company, move to AI, start a business or earn money with Ryze: the home page points you to the right universe.
Sources
- Swiss Federal Act on Unfair Competition (UCA, SR 241), Article 2: fedlex.admin.ch
- TikTok, help page on AI-generated content: support.tiktok.com
- YouTube, disclosing altered or synthetic content: support.google.com/youtube
- Meta, approach to labelling AI-generated content: about.fb.com
The platform rules were reread on 6 October 2026 and can change. This article gives general guidance, not legal advice.
Last updated: 7 October 2026 (translation).