How to Disclose an AI Influencer Endorsement

How to Disclose an AI Influencer or Virtual Endorsement

An AI influencer campaign may need at least two clear explanations: what the audience is seeing and what material relationship connects the persona to the brand. A platform's synthetic-media label does not automatically disclose sponsorship, and an advertisement label does not explain who or what is speaking.

This guide identifies questions for counsel and campaign reviewers. It is not legal advice or approval to publish a campaign.

flowchart LR
    A["Define the persona and production"] --> B["Identify operator and represented identity"]
    B --> C["Map brand and material connections"]
    C --> D["Review claims and substantiation"]
    D --> E["Design in-content disclosure"]
    E --> F["Apply current platform controls"]
    F --> G["Test placement and comprehension"]
    G --> H["Preserve, monitor, correct, withdraw"]

Document the speaker before writing disclosure copy

The review record should identify whether the audience sees a fictional character, a real person represented by a digital double, an avatar, a generated face, a synthetic voice, edited media, or a mixed production.

It should also name who operates the account, who approves posts, what software generates, what can publish automatically, who owns the character, and who is authorized to negotiate on its behalf.

"AI influencer" does not answer those questions. [[AI Influencer Virtual Avatar or Synthetic Persona]] provides the taxonomy.

Rights review belongs at the beginning. A production team may need permission for likeness, voice, performance, source assets, music, trademarks, character rights, and training or reference material. Disclosure does not cure missing authorization.

YouTube's impersonation policy makes that boundary explicit for its platform: altered or synthetic disclosure is not permission to impersonate a person, entity, or channel.

Map every material connection

In the United States, the Federal Trade Commission's Disclosures 101 describes a material connection as including financial, employment, personal, family, free-product, discounted-product, and other valuable relationships that may affect how an audience weighs an endorsement.

A virtual persona can have several relationships at once. The brand may own the character. An agency may operate it. A software company may provide free tools. A retailer may pay commission through an affiliate link. A performer may voice the character. Each relationship should be reviewed for whether it is unexpected and material to the audience.

The FTC's 2023 revised Endorsement Guides announcement says the revised definition of endorsement clarifies coverage of virtual influencers. It also notes that a platform's built-in disclosure tool might not be adequate.

Do not use a character biography as a substitute for campaign disclosure. The audience may encounter one recommended post, embedded video, screenshot, or repost without visiting the profile.

Review what the persona is claiming

An endorsement must be honest and not misleading. The FTC's Endorsement Guides questions and answers explains that an endorsement cannot make a claim the marketer could not legally make.

That creates a special problem for fictional experience. A synthetic character cannot literally have a human sensory experience unless the communication makes the representational basis clear. "I used this for thirty days" cannot be placed in a fictional mouth without examining what the audience would reasonably understand and what evidence supports the statement.

Review express claims, implied claims, demonstrations, comparisons, testimonials, health or performance claims, environmental claims, prices, limitations, and the basis of any represented experience. Qualified advertising and regulatory counsel should determine what substantiation and wording are required.

Children, health, finance, safety, elections, and vulnerable audiences can introduce additional rules and risk. Do not rely on this general guide for those campaigns.

Keep identity and sponsorship disclosures separate enough to understand

The final communication may need to explain the character, production method, operator, represented identity, sponsor, and claim basis.

These examples are issue-spotting prompts, not approved wording. The correct language depends on the actual facts, audience, claim, medium, jurisdiction, and platform.

Avoid vague labels such as "AI-powered," "collab," or an unexplained icon when they leave the essential relationship unclear.

Put disclosure where the endorsement is encountered

FTC staff guidance says disclosure should be hard to miss and placed with the endorsement. It warns against hiding disclosure at the end, behind "more," inside a group of hashtags, or only on a profile page.

For video, the FTC says the disclosure should appear in the video, not only in the description, and notes that both audio and visual presentation may improve notice. For livestreams, it says disclosure should repeat periodically because viewers enter at different times.

The same design principle applies to synthetic identity. If a realistic persona appears in a short video, disclosure in an account biography may not travel with the content. Review placement in the frame, caption, audio, expanded description, landing page, and any paid-placement wrapper.

Test the mobile experience, muted playback, small screens, fast scrolling, accessibility tools, translations, clips, and embeds. A disclosure that technically exists but cannot be noticed or understood is a weak control.

Use platform controls without treating them as the whole answer

YouTube's altered or synthetic content guidance requires disclosure for specified realistic, meaningfully altered or generated content. YouTube says the setting can produce a label in the expanded description and that sensitive topics may receive a more prominent label.

The same page says minor or unrealistic edits may not require that creator disclosure under its rule. That platform boundary is not a legal conclusion, rights clearance, or endorsement review.

Use every current platform field that applies, then add the clear in-content disclosure needed for the audience and campaign. Recheck the platform immediately before release because controls, labels, placement, and policies can change.

Preserve a campaign record

Keep the approved persona description, ownership and rights records, material-connection map, claim substantiation, disclosure copy, placement decisions, platform settings, source assets, Content Credentials where used, publication versions, screenshots or renders, translations, accessibility checks, review owners, and approval dates.

The C2PA 2.4 Content Credentials explainer describes a tamper-evident way to record assertions about origin, modification, and AI use. Credentials can strengthen the production record, but the specification says provenance alone does not determine whether depicted content is true.

Define who monitors live posts, paid placements, affiliates, automated replies, reused assets, and reposts. Establish correction and withdrawal procedures. If the persona, operator, sponsor, claim, audience, medium, or platform changes, reopen review.

Treat disclosure as an operating system

Clear wording at publication time is only one control. The campaign also needs accurate production records, rights, claim substantiation, human accountability, platform compliance, monitoring, correction, and a way to stop automation.

E023 began with Dalton's concern that synthetic influencers can arrive in a feed without enough context. The durable answer is not one universal "AI" tag. It is a disclosure system that follows the content and tells the audience the facts needed to interpret both the speaker and the commercial message.

E063 addresses voice authorization in [[How to Create a Consent Agreement for an AI Voice]]. E070 focuses on durable publishing practice in [[How to Design a Useful Synthetic Media Disclosure]]. E093 extends the brand question in [[How Brands Should Disclose Synthetic People]].

This compliance guide was developed with AI assistance from the preserved E023 captions, current FTC guidance, YouTube policies, the C2PA 2.4 explainer, and the linked disclosure record. Dalton Anderson remains the author. It is not legal advice, campaign approval, rights clearance, claim substantiation, or a universal disclosure template. Editorial, platform, identity, rights, advertising, privacy, jurisdiction-specific legal, source, accessibility, and founder review are required before publication or use. Publication is not authorized.