A virtual or AI influencer has to disclose a sponsorship exactly as a human creator does, and usually needs to say more. The US Federal Trade Commission's Endorsement Guides cover endorsers who only appear to be individuals, the UK Advertising Standards Authority counts any virtually produced persona as an influencer, and Australia's AANA Code requires advertising to be clearly distinguishable. Because a synthetic persona cannot use a product, its captions also must not claim personal experience, and in the EU content that counts as a deep fake must be labelled as artificially generated under rules applying from 2 August 2026.
This guide is for people who run an AI persona, creators who use an AI-generated version of themselves in sponsored content, and the managers who brief them. It sticks to what changes when the endorser is synthetic. The mechanics of each platform's branded-content tool live in our guides to the Instagram paid partnership label and the TikTok paid partnership setting.
Three things a synthetic endorser may need to disclose
Human creators mostly think about one disclosure, the ad label. An AI persona can trigger three separate duties, and fixing one does not fix the others.
| What must be clear | When it applies | Where the duty comes from |
|---|---|---|
| The post is advertising | Any payment, free product, commission or other connection to the brand | FTC Endorsement Guides, UK consumer law and the CAP Code, AANA Code section 2.7 |
| The persona or scene is generated | When viewers could otherwise believe a real person is speaking or real results are shown, and for EU deep fakes | ASA guidance on AI disclosure, EU AI Act Article 50, platform AI labels |
| The persona has not used the product | Whenever the script would otherwise imply first-hand use or results | FTC bona fide user rule, CMA genuine-experience guidance, ACCC misleading-conduct examples |
What each regulator says about synthetic endorsers
United States. The FTC Guides Concerning Use of Endorsements and Testimonials in Advertising say the endorser “could be or appear to be” an individual, group or institution. In the 2023 Federal Register notice adopting that wording, the Commission said the new wording was neither ambiguous nor limited to virtual influencers, because it was also meant to reach fake-review writers and non-existent entities. Read together, a virtual influencer is an endorser, and the usual duty to disclose unexpected material connections applies.
United Kingdom. The ASA's advice on recognising ads in social media and influencer marketing says it typically defines an influencer as any human, animal or virtually produced persona active on a social platform, and expects a prominent “Ad” label upfront. Its 2025 article on disclosing AI in advertising says there is no blanket UK legal requirement to disclose AI use in ads, that a disclosure cannot rescue a fundamentally misleading message, and that making clear an influencer is AI-generated could help negate an otherwise misleading impression. The CMA's guidance for content creators adds that incentivised content must be identifiable as an ad and reflect your genuine experience, and warns that hidden ads can mean a substantial fine under consumer law.
Australia. Section 2.7 of the AANA Code of Ethics, now published on the Marketing Association of Australia site, says advertising must be clearly distinguishable, and its practice note requires paid or gifted influencer relationships to be clear, obvious and upfront. The Code does not mention virtual personas; it applies to marketing communications whatever produced them and calls itself technology neutral. Under the Australian Consumer Law, the ACCC's page on social media promotions gives an example that maps neatly onto AI personas: an influencer paid to say she loved a resort she never visited is likely to be misleading.
European Union. Article 50 of the AI Act requires deployers of AI systems that generate or manipulate image, audio or video content constituting a deep fake to disclose that it is artificially generated or manipulated. The Commission's guidelines page on the transparency obligations confirms the article applies from 2 August 2026. The Act defines a deep fake as content resembling existing persons, objects, places, entities or events that would falsely appear authentic, so whether a wholly invented character qualifies is a question for EU counsel. The disclosure must be clear and distinguishable no later than the first exposure, and for evidently artistic, creative or fictional work it can be lighter, provided it does not spoil the work.
Disclosure checklist by market
Use the row for every market your audience is in, not just where you live. Where audiences overlap, meet the strictest row.
| Market | Commercial label | Synthetic persona note | Experience claims |
|---|---|---|---|
| United States (FTC) | Plain words such as “ad” or “sponsored”, placed with the endorsement, inside the video, repeated during live streams, in the endorsement's language | No standalone AI-label rule in the Guides; disclose the persona where its absence would mislead about who is speaking | No claims of personal use; product claims need the advertiser's substantiation |
| United Kingdom (ASA, CMA) | “Ad” at the start, visible on every device, not buried in hashtags or behind “see more” | No blanket legal duty, but saying the persona is AI-generated may help prevent a misleading impression | Content must reflect genuine experience; an AI disclosure does not excuse a misleading image of results |
| Australia (AANA, ACCC) | Clear upfront terms such as #ad, Advertising, Paid Partnership; weak tags like #sp, Spon, gifted or Collab may fall short | Not addressed in the Code; misleading conduct rules still apply to who appears to be speaking | Praise of an experience the endorser never had is likely to mislead |
| European Union (AI Act, plus national ad law) | Follow the national advertising rules of each member state you target | Label content that counts as a deep fake as artificially generated, at first exposure, from 2 August 2026 | National consumer law applies; get local advice before running claims-led campaigns |
The experience problem: what a synthetic persona can honestly say
This is where AI influencer briefs most often go wrong. The FTC Guides say that when an ad represents that the endorser uses the product, the endorser must have been a bona fide user, and the FTC's Disclosures 101 for Social Media Influencers puts it plainly: you cannot talk about your experience with a product you have not tried. A generated character has not tried anything. The ASA gives a parallel warning: an AI-generated image of a cosmetic effect that does not reflect real-world results is likely to mislead, and saying AI was used does not cure it.
| Risky line in a persona script | Why it is a problem | A version the persona can say |
|---|---|---|
| “This moisturiser fixed my dry skin in a week.” | Implies personal use and a result no synthetic character can have | “[Brand] says this moisturiser is made for dry skin. I'm a digital character, so I can't test it myself.” |
| “I run in these shoes every morning.” | A false claim of ongoing use | “Here's what's new in [Brand]'s running range, from the brand's own spec sheet.” |
| A generated before-and-after picture | Shows results that never happened to anyone | Real product photography supplied by the brand, with any results claim backed by the brand's evidence |
| “My favourite hotel in Bali.” | The ACCC resort example: praise for a stay that never took place | “Our team visited [hotel], paid for by [Brand], and here is what they found.” Only if a real person did |
If a real member of the team did use the product, attribute the experience to that person by name or role rather than to the character. The honest opinion must belong to whoever is said to hold it.
AI-generated reviews, testimonials and lookalikes
The FTC's rule on consumer reviews and testimonials, announced in a 2024 press release, prohibits reviews and testimonials that misrepresent that they come from someone who does not exist, naming AI-generated fake reviews as an example. The FTC's questions and answers on the rule draw the line carefully: there is no blanket ban on AI avatars in marketing, the relevant section was drafted so as not to prohibit virtual influencers, and an avatar's testimonial breaks the rule only if the underlying testimonial is fake or false. The same answer warns that using a celebrity avatar without permission breaks the rule where viewers would think the celebrity actually gave the testimonial.
Two further points apply to lookalikes. The Endorsement Guides say using an endorsement with the image or likeness of someone other than the actual endorser is deceptive if it misrepresents a material attribute of the endorser. In the UK, the ASA's 2026 note on AI and deepfakes says an AI-generated celebrity endorsement realistic enough to mislead is likely to breach the Code, and that personality and intellectual property rights, which the ASA does not regulate, still need legal advice. If you license a real person's face or voice for an AI twin, our digital twin licence checklist covers the contract side.
Caption templates for AI persona sponsorships
Replace the bracketed parts and keep the label at the start. Write the caption in the same language as the post, as the FTC asks, and keep claims to what the brand can back up.
- Paid feed post or Reel: “Ad | Paid partnership with [Brand]. I'm an AI-generated character run by [studio or operator]. I can't try products, so here's what [Brand] says about [product]: [substantiated claim].”
- Gifted product tested by a human: “Ad: [Brand] gifted this to the team behind this account. I'm a virtual creator; [name or role], a real person, used it for [period] and thought [specific finding].”
- Affiliate link: “Ad: this account earns commission on purchases through this link. Virtual influencer, operated by [studio].”
- Short video: on-screen text in the opening frames reading “Ad for [Brand] | AI-generated host”, also spoken aloud, because the FTC notes some viewers watch muted and others miss overlays.
- Live stream: a pinned line plus a spoken reminder at regular intervals: “This stream is sponsored by [Brand], and I'm a digital host.”
- A human creator's own AI double: “Ad for [Brand]. This video uses an AI-generated version of me. I approved the script, and I do use [product].” Keep the last sentence only if it is true.
Labels to avoid on their own, drawn from the FTC and AANA guidance: “sp”, “spon”, “collab”, “gifted”, “thanks to [Brand]”, “ambassador” with no brand name, or a brand tag with no label at all. Put the persona's AI status in your media kit too, so brands brief you with it in mind.
Who is responsible when a persona gets it wrong
The FTC Guides spread liability widely. Advertisers can be liable for misleading endorsements and missing disclosures, and should guide, monitor and correct their endorsers. Endorsers can be liable for their own statements. Advertising agencies, PR firms and similar intermediaries can be liable for their roles in creating or spreading deceptive endorsements, including by directing endorsers who fail to disclose. For an AI persona, the person or studio publishing the character stands in the endorser's shoes; there is no one else to blame.
The UK picture is similar. The ASA says that when an advertorial breaches the CAP Code, both the influencer and the brand are named in its ruling, and that brands are usually treated as at least jointly responsible for affiliate content. Put the disclosure wording, the AI note and the approval process in the written brief so everyone signs off on the same text.
Platform tools sit on top of the law
The FTC tells influencers not to assume a platform's disclosure tool is good enough, while encouraging them to use it alongside their own disclosure, and the CMA accepts platform labels such as “paid partnership” where they are clear. AI labels work the same way: they inform viewers but do not replace an advertising disclosure. On YouTube, for example, realistic synthetic content needs the AI use answer in Studio as well as any paid promotion setting; our YouTube AI disclosure guide covers when that applies. Persona accounts on subscription platforms face their own AI rules too, such as those in our Fanvue AI creator checklist.
Limitations of this guide
This is general information, not legal advice, based on regulator and code pages as they read on 1 October 2026. It covers four markets only, leaves out sector rules for products such as alcohol, gambling, finance and health that add their own restrictions, and does not decide whether a particular invented character is a deep fake under EU law. Self-regulatory codes and regulator guidance are applied case by case, so following these templates reduces risk without removing it. For campaigns that make health or results claims, use real people's likenesses or target several countries, get advice from a lawyer qualified in each market, and for UK copy the ASA's Copy Advice team offers a pre-publication check.