Virtual Influencer Ethics — The Boundaries of an AI-Made "Persona"— When a generated narrator speaks about medicine, how far is permitted?

01What a Virtual Influencer Is — Defining the Term

A virtual influencer (= a communicator that does not exist in reality, built with CG or generative AI) is a "persona" whose photos, videos, and text are all artificially produced. No living human is performing the role; the appearance, the voice, and the words of the posts are all designed by the creator. Lil Miquela, which appeared around 2016, is the representative example — she has signed contracts with real brands and introduced products just like an ordinary influencer.

Technically, three layers combine. Understanding them separately makes the later regulatory discussion easier to organize.

• Generating the appearance — 3DCG or image-generation AI produces a consistent face, body, and expressions. The same "person" can be generated in hundreds of images across different backgrounds and outfits.
• Generating the words — Posts and comment replies are drafted by a large language model (= an AI trained on vast amounts of text). Tone of voice and even values are part of the design.
• Designing the operation — Which product, in what context, with what emotion, gets recommended. This "editorial policy" is decided, in the end, by the humans or companies behind the scenes.

The crucial layer is the last one. A virtual influencer may look like "AI talking on its own," but in reality there is always a party responsible for the communication (the creator, the advertiser) standing behind it. This fact is the starting point when considering whether such a figure may be used in pharmaceutical advertising.

02Why They Are Spreading, and Real-World Examples

Why are fictional communicators used in advertising? The reason is clear: the creator can bring the uncertainty that clings to a living influencer entirely under control.

In consumer goods, use has spread mainly around beverages, cosmetics, and apparel. Some companies own their own virtual characters and communicate through official accounts. In pharmaceuticals and healthcare, by contrast, having such communicators recommend products is almost never seen in Japan. This is not self-restraint but an institutional wall. From the next chapter, we verify that wall against the regulatory text.

03Whether It Is Permitted in Medical / Pharmaceutical Advertising — The Conclusion First

Let's state the conclusion up front. Having a fictional communicator recommend a prescription drug to the general public is not permissible under Japan's framework. The reason is not singular; several regulations overlap and point to the same conclusion.

Organized, it forms a three-tier structure.

• Prescription drugs cannot be advertised to the general public in the first place — The Standards for Appropriate Advertising of Pharmaceuticals do not, in principle, permit advertising of prescription drugs to the general public. This is a matter that precedes any argument over whether the communicator is real or fictional.
• If it qualifies as advertising, the Pharmaceutical Act applies in full — Whoever the communicator is, if the post is judged to be "advertising," the prohibition on exaggerated representation (Pharmaceutical Act Article 66) and the prohibition on advertising unapproved drugs (Article 68) apply.
• Faking the communicator triggers the stealth-marketing rule under the Premiums and Representations Act — Concealing that something is advertising and disguising it as a third party's impression violates the 2023 notification discussed below.

For OTC drugs (= over-the-counter medicines that can be bought at a pharmacy) or health-related consumer goods, advertising to the general public is itself possible. Even then, using a fictional communicator raises a new transparency problem: "whose experience is this?" We examine this point in detail in Chapters 05 and 06.

04Advertising Applicability Under the Pharmaceutical Act — Articles 66, 68, and 68-2

Here we pin down the text of the Pharmaceutical Act precisely. Because getting the article numbers wrong throws off the whole discussion, we confirm them carefully.

• Article 66 (Prohibition of exaggerated advertising, etc.) — One must not advertise, explicitly or implicitly, false or exaggerated articles regarding the name, method of manufacture, efficacy, effect, or performance of a drug. Guaranteeing an effect also counts as exaggerated representation.
• Article 68 (Prohibition of advertising unapproved drugs, etc.) — One must not advertise the name or efficacy of a drug, medical device, etc., that has not received approval. Having a fictional communicator voice expectations about a product still in development falls under this.
• Article 68-2 (Duty of effort to provide information) — A marketing authorization holder, etc., has a duty to endeavor to provide information on quality, efficacy, and safety to healthcare professionals. This is a framework of information provision, separate from "advertising."

To judge advertising applicability, the "three requirements" set out by the authorities are used. (1) the intent to induce customers is clear, (2) the product name of a specific drug, etc., is made clear, and (3) it is in a state recognizable by the general public — when all three are satisfied, it is judged to be "advertising." A virtual influencer's post, if it names a product, prompts a purchase, and is viewable by anyone through a public account, easily satisfies the three requirements. In other words, the excuse that "it is just an AI-made character's monologue, so it is not advertising" does not hold.

05The Stealth Marketing Rule — The 2023 Premiums and Representations Act Notification

On October 1, 2023, a new notification under the Premiums and Representations Act (Act against Unjustifiable Premiums and Misleading Representations) took effect — the so-called stealth marketing rule. Formally, it designates "a representation for which it is difficult for the general consumer to determine that it is a business operator's representation" as an unfair representation under Article 5, item 3 of the Act.

The core of this rule comes down to one point: it prohibits concealing that something is a business operator's advertising and making it appear to be a third party's voluntary impression. In relation to virtual influencers, this carries a double meaning.

Organized, using a fictional communicator is not itself immediately illegal. What is prohibited is "concealing that it is a business operator's advertising." Therefore, if both the fictional nature and the advertising nature are disclosed, there is room for it to stand in the consumer-goods domain. The question is how thoroughly that disclosure can be carried out in a pharmaceutical context — which moves us to the next chapter's issue.

06The Craft of Transparency Disclosure — What to Write, Where, and How

When using a fictional communicator, securing transparency is not a matter of "adding one line of note." The following elements must be delivered in a form the consumer can actually perceive.

• Disclosure that it is advertising — Place "Ad," "PR," "Promotion," etc., at the start of the post or in a visible position. Tucking it into a trailing hashtag tends to be judged as hard to determine.
• Disclosure that it is a fictional being — Show, in a discernible form, that the communicator is not a real person but a generated character. It should be recognizable not only in the profile field but in each individual post.
• Avoid feigning experience — Do not have it speak of the "felt use" of a product it has not used. Asserting effects or feigning personal experience can violate Article 66 for drugs and the stealth-marketing rule for consumer goods alike.

Here a heavy pharma-specific constraint is added. Sales-information-provision activities for prescription drugs are, under the "Guideline on Sales Information Provision Activities for Prescription Drugs" (= the "HanteiG," notification of the Director-General of the Pharmaceutical Safety and Environmental Health Bureau, MHLW, 2018), required to observe the distinction between information provision and advertising, disclosure of supporting materials, and clear identification of the responsible person's name and affiliation. The premise is that the recipient can verify who delivered the information. A fictional communicator structurally struggles to satisfy this "being verifiable." Here lies the fundamental incompatibility between prescription drugs and a generated narrator.

07Collision With a Credence Good — Why It Is Especially Strict for Pharmaceuticals

Since Vol. 1, this series has treated pharmaceuticals as a "credence good" — a good whose quality the buyer can hardly verify on their own either before or after purchase, and for which they must depend on trust in the communicator. This property collides head-on with the fictional communicator.

The collision happens on two layers.

• Verifiability of the information source — For life-critical information, "who spoke, with what qualifications and what evidence" matters. A fictional communicator cannot answer that question in substance. Feigning an expert's recommendation, on top of the Pharmaceutical Act problem, undermines trust itself.
• The danger of emotional design — A virtual influencer's strength lies in being able to deliberately design intimacy and empathy. But for pharmaceuticals, that designed intimacy creates the danger of steering a patient's judgment toward excessive expectation of an effect.

International principles support this direction as well. The World Health Organization's "Ethical Criteria for Medicinal Drug Promotion" (1988) holds that drug promotion should be accurate, fair, and verifiable, and cautions against misleading representations and excessive appeals to emotion. Before asking whether the communicator is real, it is a principle that requires the substance of the promotion to be verifiable and honest. A fictional narrator struggles to meet this verifiability requirement.

08Design Guidance — For Judgments in the Pharmaceutical Field

So, when marketing and medical staff at pharmaceutical companies hesitate in practice, how should they draw the line? We organize it by domain.

Three common axes of judgment apply. First, confirm at the outset whether the post satisfies the three requirements for advertising applicability. Second, if it does, examine the constraints of the Pharmaceutical Act (exaggeration under Article 66, unapproved products under Article 68) and the appropriate-advertising standards side by side. Third, even when it falls into the consumer-goods domain, ask yourself from the stealth-marketing angle whether anything is being "concealed." When in doubt, return first to the three requirements for advertising applicability — this is the safe first move in the field.

09Connections to Other Chapters on This Site

This installment's issues connect to other chapters, inside and outside the series, as follows. Reading them together adds depth to how the boundary is drawn.

• AI Marketing Vol. 2 — Vibe Marketing — The "feel" design that a fictional communicator carries, as a technique for building a brand's atmosphere.
• AI Marketing Vol. 5 — Generated Content and Review — The mechanism that runs generated output through regulatory checks. A fictional communicator's posts are subject to the same review.
• Ad Regulations Vol. 1 — Pharmaceutical Act §§66–68 — A detailed reading of the articles used in this installment.
• Material Review series — The practice of who reviews a generated communicator's posts, and how.

The virtual influencer is a technology that gives a "persona" to a narrator who does not exist. In consumer-goods branding, it has grown into a leading player in advertising, armed with control over scandal risk and a consistent worldview. But when one tries to bring it into the prescription-drug domain, three walls rise at once: the appropriate-advertising standards' in-principle ban on advertising to the general public, the Pharmaceutical Act's prohibition of exaggerated and unapproved representations under Articles 66 and 68, and the 2023 stealth-marketing rule. Deeper still lies the property that pharmaceuticals are a credence good — the weight of the recipient being able to verify who spoke, and on what evidence. A fictional narrator structurally struggles to meet this verifiability. So even though it is usable as a technology, broad domains remain in which it cannot be used for pharmaceutical product appeal. When judgment falters, begin the examination not from whether the communicator is real or fictional, but from whether the post qualifies as "advertising." The boundary can be drawn from there. Next time, we go into the implementation of putting AI into customer relationship management — and its pitfalls.