AI Avatars, Testimonials, and Product Claims: What Brands Need to Know

AI makes it possible to create a convincing presenter without a studio, translate a script into several languages, and produce advertising scenes without organizing a traditional shoot.

But a convincing presentation is not the same as a truthful one.

A virtual presenter explaining your product is different from a fictional customer claiming to have used it. A genuine review is different from an invented success story. And access to a music track or synthetic voice does not automatically mean you have permission to use it in paid advertising.

The guiding principle is truthfulness: be accurate about who is speaking, what they experienced, what the product can do, and which rights you have secured.

This article provides a practical framework, drawing primarily on U.S. FTC guidance. It is not legal advice. Requirements vary by jurisdiction, product category, platform, and the details of the campaign.

A Virtual Presenter Is Not Automatically a Customer

The first question is not whether the person on screen is AI-generated.

It is what viewers are being led to believe about that person.

Consider these two scripts:

"This platform brings approval requests into one workspace. Here is how the workflow operates."

"I switched my team to this platform last month, and we stopped missing approvals."

The first presents product information. The second claims a personal experience.

If the second speaker is a fictional customer whose experience never happened, the issue is not simply the use of AI. It is the fabricated testimonial.

FTC staff guidance explicitly states that the Consumer Reviews and Testimonials Rule does not impose a blanket prohibition on AI-generated avatars. However, an avatar's testimonial may violate the rule if the underlying experience is fake or false. The presentation may also be deceptive under the FTC Act.

Define the presenter's role

Before production, decide whether the person on screen is:

• A brand presenter explaining information.
• A real founder speaking about their own business.
• A real customer describing an actual experience.
• An actor or avatar portraying someone else.
• A character in an obviously fictional dramatization.

These roles require different handling. (When each role fits your brand best is covered in Will AI-Generated Ads Make Your Brand Look Cheap?)

A fictional scene can illustrate a problem without pretending to document a customer's experience. But presentation matters: names, job titles, customer labels, first-person narration, and interview-style editing can all influence what viewers believe.

Do not rely on your internal description of the video. Evaluate the impression the finished ad creates.

Why You Cannot Invent Product Experience

A testimonial carries an implication: someone actually encountered the product and is accurately describing what happened.

That implication gives it value. It is also why inventing the experience is problematic.

The FTC's Endorsement Guides explain that endorsements must reflect honest opinions and experiences. Someone who has not tried a product cannot truthfully describe using it, and someone who tried it once cannot imply regular use.

Avoid scripts such as:

• "I have used this every morning for six months," when no such user exists.
• "Our agency doubled its output," attributed to a fabricated agency owner.
• "This cleared my skin in seven days," presented as an invented customer account.
• "My team switched last week," delivered as an experience that never occurred.

Calling the presenter "AI-generated" does not, by itself, make the claimed experience true.

Instead, change the communication approach:

Unsupported approach	More appropriate alternative
Invented customer success story	A clearly framed illustrative use case
Fictional first-person product experience	A presenter explaining verified product capabilities
Fabricated performance result	A supported claim with relevant context
Synthetic customer interview	An authorized account from an actual customer

A dramatization should be clearly recognizable as such. It should not function as a disguised testimonial.

Work with Real Reviews Without Changing Their Meaning

Genuine customer feedback can be valuable source material. Turning it into advertising requires care.

FTC staff guidance notes that a consumer review featured in a company's marketing becomes a testimonial. It is no longer merely a review hosted on a website.

Use a documented workflow.

Preserve the source

Keep a record of:

• The original review or interview.
• Its date and source.
• The relevant product or service.
• The reviewer's actual wording.
• Any known incentive or relationship.

This creates a reference for checking the final script.

Edit for length, not a different conclusion

Shortening a quotation should not change its meaning.

For example:

"Setup took some work, but once we connected our sources, monthly reporting became easier."

should not become:

"Setup was easy, and reporting takes no work."

The FTC's guidance cautions against presenting endorsements out of context or editing reviews in ways that misrepresent customers' views.

Do not combine comments from several people into one quotation attributed to a single customer.

Secure appropriate permission

As a production practice, obtain written authorization for the uses you plan, particularly when including someone's name, image, voice, company identity, photographs, or recordings.

Make the intended scope explicit:

• Paid advertising or organic content.
• Platforms and territories.
• Duration of use.
• Editing and translation.
• Any synthetic voice or likeness treatment.
• Agency or production-partner involvement.

Do not assume that a publicly visible review authorizes every advertising use. The legal and contractual position can depend on the source, jurisdiction, and material involved.

Be especially careful with synthetic delivery

A real quotation read by an avatar can still mislead if viewers believe the avatar is the actual customer.

FTC guidance explains that actors portraying testimonialists can create deception concerns even when the underlying testimonial is genuine.

If considering a reenactment, translation, or synthetic reading, review the permissions and the overall presentation. Clearly distinguish the actual source from the person—or avatar—delivering it.

Disclose Relevant Relationships

A real endorsement can still be misleading if the audience does not understand an important relationship behind it.

FTC guidance calls for clear and conspicuous disclosure of unexpected material connections that could affect how people evaluate the endorsement. These can include payment, free products, employment, and certain personal or business relationships.

Ask:

• Was the person paid?
• Did they receive a free or discounted product?
• Are they an employee, founder, relative, or business partner?
• Would the audience otherwise understand that connection?

Do not assume a disclosure is adequate merely because it exists somewhere in the caption or a platform tool has been activated.

FTC staff warns that built-in disclosure tools may be insufficient when the resulting notice is easy to miss, fleeting, or difficult to read.

Treat Customer Statements as Advertising Claims

"Those were the customer's words" is not a complete defense for using an unsupported claim.

An endorsement cannot communicate a claim the advertiser could not lawfully make directly. Advertisers can be responsible for misleading or unsubstantiated statements conveyed through endorsements.

Review both the quotation and the broader impression.

For example:

• Does one customer's result appear to promise the same result to everyone?
• Does the edit remove conditions that affected the outcome?
• Does a before-and-after sequence imply a causal effect?
• Does narration add a numerical claim that the customer never made?

A truthful exceptional result may still mislead if presented as representative. FTC guidance addresses the need to substantiate generally expected results and, where appropriate, clearly disclose what consumers can generally expect.

For every objective claim, record:

Item	What to document
Claim	The exact statement or implied message
Evidence	What supports it
Scope	Which products, users, or conditions it covers
Limitations	Relevant exclusions or qualifications
Approval	Who confirmed it is suitable for use

Apply this review to visuals as well as words. A generated demonstration can imply product performance even without an explicit promise — see what to generate and what to keep real in e-commerce ads.

Check Licenses Across the Entire Asset Chain

A paid subscription is not a substitute for checking usage rights. (Rights and licensing are also a line item worth asking about when you work out what an AI video ad really costs.)

Before publishing, review the licenses for the inputs, tools, and outputs involved in the ad.

Music

Confirm whether the license covers:

• Paid advertising.
• Your intended platforms.
• Relevant territories and campaign duration.
• Editing and synchronization with video.
• Use by your agency or on behalf of a client.

Do not assume music available inside one platform can be exported for advertising everywhere else.

TikTok's commercial terms subject Commercial Music Library use to separate user terms. Its advertising terms also require advertisers using music from outside that library to obtain the necessary rights, licenses, and permissions.

Voices and likenesses

Check permission for the intended use of a person's voice or appearance, including any cloning, synthesis, translation, or digital-replica treatment.

Confirm that the tool's license covers the selected stock voice or avatar and the planned advertising use.

Do not treat technical access to someone's voice or photograph as consent to create an endorsement. FTC guidance specifically identifies unauthorized celebrity avatars as a potential rule violation when consumers would believe the celebrity actually gave the testimonial.

Images, footage, and graphics

Review commercial-use permissions, editing restrictions, and any relevant model or property releases.

Check assets supplied by contractors as well as those your team downloads directly. "The freelancer found it" is not a rights record.

Generative tools

Read the terms applicable to the plan and features you actually use.

Check:

• Whether commercial advertising is permitted.
• Restrictions on people, brands, and product categories.
• Your responsibility for uploaded reference material.
• Rules governing output use.
• Any watermark or attribution conditions.
• Confidentiality and data-use provisions.
• Whether client work or asset transfer is covered.

Keep a record of the applicable terms and license documentation. Tool output should still pass your own content, accuracy, and rights review.

Check Platform Rules and the Law Separately

There is no single "AI-compliant" approval that covers every use.

Review three separate layers:

Layer	Main question
Law and regulation	Is the campaign lawful in the relevant markets?
Platform policy	Is this content permitted in this placement and category?
Contract and license	Do we have permission to use these assets this way?

Passing one review does not establish compliance with the others.

A music license does not make a false testimonial acceptable. A platform's acceptance of an ad is not legal clearance. A lawful creative approach may still be restricted by a particular platform or advertising category.

Before launch, check current requirements for synthetic-content labeling, endorsements, restricted products, and disclosures in each intended placement. Do not assume the same export can run unchanged everywhere.

Build a Review Gate Before Publishing

For a practical final check, ask:

1. Who does the viewer believe is speaking?
2. Is every claimed personal experience genuine?
3. Have quotations retained their original meaning?
4. Are permissions appropriate for the planned use?
5. Are material relationships disclosed clearly?
6. Are objective claims supported?
7. Do the visuals imply anything unsupported?
8. Are music, voice, image, and tool rights documented?
9. Have platform rules and applicable law been checked separately?
10. Has the final export been reviewed—not just the script?

Seek qualified legal review when the campaign involves sensitive claims, regulated categories, unclear permissions, or realistic synthetic portrayals of identifiable people.

To make this gate part of every batch rather than a one-off check, build it into your repeatable video ad workflow.

Use AI to Improve Production, Not Manufacture Credibility

AI can help you explain a product, illustrate a situation, and produce creative versions more efficiently.

It should not be used to invent the independent experience that makes a testimonial persuasive.

At ShotCode, the practical distinction is straightforward: use presenters to present, authentic customer accounts to describe real experiences, and supported evidence to communicate product performance.

That separation leaves plenty of room for creative advertising—without asking viewers to trust a story that never happened.

For the practical side, see Can You Create AI UGC Ads Without Fake Testimonials?, How to Turn Real Customer Reviews into Video Ads, and Do AI Video Ads Need a Disclosure Label?.