Ethics of influence: AI, virtual influencers, deepfakes and edits

Generative artificial intelligence has made it cheap and easy to create realistic images, voices and videos, to edit appearance and results invisibly and to build entire personas that do not exist. These tools can help creators work faster and more creatively. They also make deception easier than ever: a product can be shown doing what it cannot do, a person can appear to endorse something they never saw, and an audience can form a relationship with a character it believes is human. This lesson looks at the ethics and law of influence in this new environment: edited and filtered images, AI-generated content, virtual influencers, deepfakes and the rights people have over their names, faces and voices. It ends with a practical framework for ethical decisions.

Ethics as a professional practice

Lesson 1 introduced a four-question ethics check: honesty, fairness, respect and accountability. Professional ethics goes further than asking those questions occasionally. It means building habits and policies that make the right choice the default, especially under commercial pressure. Creators face real tension: brands want results, audiences reward polish, and competitors may cut corners. Ethical creators handle this by deciding their standards in advance, writing them down and applying them consistently, rather than deciding case by case when money is on the table.

• Decide in advance which categories and claims you will not promote.
• Write a short public or private standards statement covering disclosure, editing and AI.
• Build standards into contracts so brands know them before work begins.
• Review your choices periodically, especially after mistakes or audience feedback.

Edited images and filters

Editing is part of photography and video, and much of it is harmless: adjusting brightness, cropping, removing a stray hair. The ethical and legal question is whether an edit changes the audience's understanding of what the product does or what a person looks like in a way that matters to their decision.

• Acceptable in most cases: lighting and color correction, cropping, stabilization and background clean-up that do not affect the product's performance.
• Risky: smoothing skin, whitening teeth, slimming bodies or lengthening lashes in content promoting skincare, cosmetics, toothpaste, fitness or mascara, because the edit can exaggerate the product's effect.
• Misleading: before-and-after images where the "after" has been edited, or demonstrations sped up or altered to show results the product does not achieve.

In 2021 the UK's Advertising Standards Authority ruled that influencers should not apply filters to images promoting beauty products if the filter is likely to exaggerate the product's effect, and some countries, such as France and Norway, require labels on commercially retouched images of bodies. Canada and the United States do not have equivalent labeling laws, but their general rules on misleading advertising lead to the same practical conclusion: do not use edits that exaggerate a product's effect, and a disclosure cannot fix an image that misleads.

AI-generated content in sponsored posts

AI tools can generate product images, backgrounds, voiceovers, scripts and entire videos. Using AI is not unlawful in itself. The legal and ethical issues arise in three situations.

• When AI shows a product doing something it cannot do, or achieving results it does not achieve, the content is misleading, whether or not it is labeled as AI.
• When AI creates a person, voice or experience the audience believes is real, such as a "customer" testimonial that no customer gave, the content is deceptive.
• When AI uses someone's likeness or voice without permission, it may infringe their rights, discussed below.

The FTC's 2023 Endorsement Guides define endorsers to include virtual influencers, so their endorsements must reflect honest representations and material connections must be disclosed. The Consumer Reviews and Testimonials Rule specifically prohibits AI-generated fake reviews. The FTC has also warned repeatedly that claims about AI, and uses of AI to deceive, fall under Section 5; in 2024 it announced a law enforcement sweep against deceptive AI claims and schemes, and its Rule on Impersonation of Government and Businesses, in force since April 2024, allows civil penalties against those who impersonate businesses or government agencies, including with AI.

Virtual influencers

Virtual influencers are computer-generated characters with names, personalities and followings, operated by companies or creators. Some are clearly fictional, like cartoon mascots; others are designed to look and behave like real people. Brands use them because they are controllable, never age and never cause scandals of their own.

The ethical concerns are about honesty. If audiences believe a virtual influencer is a real person, they may trust its recommendations as they would a human's lived experience. A virtual influencer cannot taste a snack, feel a mattress or see results from a skincare product, so testimonials about such experiences are inherently misleading. And because virtual influencers are often owned by brands or agencies, the material connection may be total: the "influencer" is the advertiser.

• Disclose that the influencer is virtual, prominently and regularly, not only in the bio.
• Disclose who owns or operates the character, especially if it is the brand itself.
• Do not let a virtual influencer claim personal experiences it cannot have.
• Apply all ordinary disclosure rules to every sponsored post.

Deepfakes and fake endorsements

A deepfake is synthetic media in which a person's face, body or voice is generated or manipulated to show them saying or doing something they did not. Deepfake endorsements, in which celebrities or creators appear to promote products they never endorsed, are now common in scam advertising: in 2023, for example, an actor warned fans about an AI-generated video using his likeness to promote a dental plan, and a well-known creator warned about a deepfake ad offering phones for $2. Creators can be victims of deepfakes, and creators who use them, even as jokes, can cause harm and legal liability.

• Never use AI to make a real person appear to endorse a product without their written permission.
• If you find a deepfake of yourself promoting a product, report it to the platform, warn your audience, keep evidence and consider legal advice.
• Be careful with parody: satire of public figures can be lawful, but using a real person's likeness in commercial content is risky.

Rights over name, image and voice in the United States

Most states recognize a right of publicity, which protects people from unauthorized commercial use of their name, image, likeness or voice. Some states protect it by statute, such as California's Civil Code section 3344 and New York's Civil Rights Law, which since 2021 also addresses digital replicas of deceased performers; others by common law. Tennessee's ELVIS Act, in force since July 2024, specifically protects voices against unauthorized AI imitation. At the federal level, the TAKE IT DOWN Act of 2025 makes it a crime to publish non-consensual intimate imagery, including AI-generated deepfakes, and requires platforms to remove such content within 48 hours of a valid request. False endorsement claims can also be brought under Section 43(a) of the Lanham Act, and proposed federal legislation on digital replicas has been debated in Congress.

Authenticity and parasocial relationships

Influence works partly through parasocial relationships: one-sided bonds in which viewers feel they know a creator. These relationships can be positive, offering community and connection. They also create responsibility. A creator who seems like a friend carries more persuasive power than an ordinary advertiser, and audiences, especially young or lonely ones, may follow recommendations with less scrutiny. Ethical creators use this trust carefully: promoting only products they believe in, being honest about failures and dislikes, and not manufacturing intimacy, such as fake personal stories or "scarcity" pressure, to drive sales.

An ethical decision framework

When facing a hard choice, a structured approach helps. Score a proposed post or deal on each of the four questions from 0 to 2, where 0 means a clear problem and 2 means no concern, and add a fifth question on transparency of technique: would the audience feel misled about how the content was made?

A score of 9 or 10 suggests a sound choice; 7 or 8 suggests changes; anything with a 0 should be rejected or redesigned. Like the disclosure rubric in lesson 3, this is a thinking tool rather than a rule.

Responsibilities when things go wrong

Ethical practice includes how you respond to mistakes. If a sponsored post turns out to be misleading, or you used an edit you now regret, act quickly: correct or remove the content, tell your audience plainly what was wrong, tell the brand and, if a regulator or platform contacts you, cooperate. Audiences forgive honest mistakes handled openly far more readily than cover-ups discovered later.