AI Testimonials and the Fake Review Rule: What Brands Can No Longer Buy

There's a version of "AI testimonial" that's just modern ad production: you take a real customer's written review, put a face and a voice to it, and run it as creative. And there's a version that's a federal violation with a per-instance dollar penalty attached: you invent a customer who doesn't exist, write words they never said, and publish it as proof. Both can look identical on screen. The difference is entirely in what's true underneath the video, and since October 2024 that difference has had a specific rule and a specific enforcement history behind it — including a 2025 update that changes who the FTC is actually chasing, without changing what's still illegal.

Here's the rule in plain terms, the line that actually matters for brands, and where an AI content tool sits relative to both.

What the rule actually bans

The FTC's Consumer Reviews and Testimonials Rule — 16 CFR Part 465, effective October 21, 2024 — bans reviews and testimonials that misrepresent who's giving them: a review attributed to someone who doesn't exist, someone who never actually used the product, or someone whose stated experience doesn't match what they actually had. The rule doesn't carve out an AI exception, and it doesn't need one — it applies to AI-generated fake reviews the same way it applies to a human-written one someone made up. The mechanism of fabrication is irrelevant to the rule; only the fabrication itself is.

The same rule has a second, separate prohibition that catches a lot of ordinary-looking growth tactics: buying, selling, or otherwise procuring fake indicators of social media influence — purchased followers, views, likes, or reposts used to misrepresent a product's or a page's real reach. There's a knowledge threshold built in: you're not liable for followers you had no reason to suspect were fake, but you are liable once there were red flags you ignored. "The bot-follower package was suspiciously cheap and arrived overnight" is exactly the kind of red flag that threshold is built to catch.

Enforcement has teeth attached: the rule authorizes civil penalties on a per-violation basis, currently set at roughly $51,744 per violation under the rule's penalty provision. That's per fake review, not per campaign — a handful of fabricated testimonials scattered across a product page adds up fast if the FTC decides to look.

The line that actually matters

Here's the distinction that determines which side of the rule an AI-assisted ad lands on: is the AI playing a fictional character, or is it playing a specific real person?

An AI avatar delivering your brand's own product claims — "here's what this serum does," spoken by a synthetic presenter who's never pretending to be a customer — is ordinary ad creative. Nobody's deceived about who's talking, because nobody's claiming the avatar used the product. That's a spokesperson, and spokespeople have always been allowed to be actors, animated characters, or AI.

The violation shows up the moment that same avatar switches from spokesperson to witness: "I've used this for six months and my skin has never looked better," delivered by a synthetic persona with no real person behind the claim. That's not a stylistic choice about how you deliver a message — it's a fabricated first-person account of an experience nobody had, which is precisely what the rule exists to stop, whether the fabrication came from a copywriter's imagination or a language model's.

The safe pattern is simple to state and easy to build a review process around: if a real customer said it, dramatizing it is production. If nobody said it, generating it is the violation — regardless of how polished, how animated, or how obviously-synthetic the delivery is. Realism isn't the trigger here the way it is for platform disclosure rules; fabrication is.

The Rytr case, and what actually changed in 2025

The clearest signal of how seriously this gets enforced against AI tools specifically came in December 2024, when the FTC approved a final order against Rytr — a company selling an AI "testimonial and review" generator — for providing subscribers the means to produce detailed, specific-sounding reviews with no connection to the user's actual input, which the FTC alleged were near-certain to be false the moment someone copied and published one. The order barred Rytr from advertising, promoting, or selling any service dedicated to generating consumer reviews or testimonials.

That order didn't stay in place. In December 2025, the FTC reopened and set aside its own Rytr order, citing the White House's AI Action Plan and concerns that the underlying legal theory — holding a general-purpose AI tool liable simply because subscribers could misuse it — represented regulatory overreach rather than a sound reading of the FTC Act. The agency's own language was direct: condemning a technology because it potentially could be misused is a different thing from condemning the misuse itself.

What's important for a brand or an AI content service to understand is exactly how narrow that reversal is. It targets one specific enforcement theory — tool-provider liability for downstream misuse — not the underlying rule. The FTC was explicit that it will continue holding accountable "actors that use AI to violate the law or deceive consumers," and nothing in the reversal touches the Consumer Reviews and Testimonials Rule itself, which remains fully in force. A business that actually publishes a fabricated AI-generated testimonial is exactly as exposed today as it was the day the rule took effect. What changed is who absorbs the risk for a general-purpose tool being available at all — not what happens to whoever presses publish on a fake review.

What this means if you're building AI content into your ad stack

For a brand, the practical takeaway hasn't moved even though the tool-liability headline did: build your review process around the "did a real person say this" question, not around whether AI touched the production. Tools that generate fictional ad personas, brand spokespeople, or dramatized real customer quotes sit outside the rule entirely. Tools — or workflows — that manufacture first-person product experiences attributed to invented people sit squarely inside it, and the 2025 Rytr reversal doesn't create cover for the brand that publishes the output; it only narrows who else might share liability for having built the tool.

That's the design principle worth holding a testimonial-video workflow to: it should take real input — an existing written review, a case study, a quote you can point to — and turn it into video, not manufacture the underlying claim from nothing. The AI Testimonial Video Generator is built around exactly that constraint: it turns a review, case study, or customer quote you already have into a talking-head video with a presenter, a voice, and product cutaways — the AI does the production, not the testimony.

A compliant walkthrough

Here's how that actually runs for an ecommerce brand sitting on real reviews it hasn't done anything with yet:

1. Pull an actual review. A real customer quote from your storefront, a support ticket, a survey response, or a case study — something you can point to as genuinely said, ideally with the customer's permission to feature it.
2. Keep the words the customer's, not the model's. Paste the real quote in as the script rather than asking a generator to "write a testimonial about X" — that's the entire line between dramatizing evidence and fabricating it.
3. Open the AI Testimonial Video Generator and generate a presenter, voice, and product cutaways around that exact quote.
4. Caption it clearly as advertising, not as an unboxed, unsolicited customer video — the ad should read as your brand's production of a real review, which is what it is.
5. Keep the source on file. If a platform or the FTC ever asks where a testimonial came from, "here's the original review and the customer's permission" ends the conversation in a way "the model generated it" never will.

The same discipline applies whether the finished UGC-style ad is sold to a brand through a creator-for-hire arrangement or produced as owned content on your own ecommerce channels — the format doesn't change the rule. It stays honest as long as it's framed as brand advertising delivered in testimonial style, not passed off as an organic, unsolicited post from an actual stranger.

FAQ

Is it illegal to use AI to make a testimonial video?

No — using AI to produce or dramatize a testimonial is fine. What's illegal is fabricating the underlying testimony: a review or quote attributed to a customer who doesn't exist, or who never actually said what's being claimed. The AI is a production tool either way; the violation is in the content, not the tool.

Can an AI avatar deliver a customer testimonial?

Only if a real customer's actual words are behind it. An AI avatar reading a genuine, sourced review is defensible ad production. An AI avatar improvising a first-person product experience with no real customer behind it is a fabricated testimonial regardless of how it's delivered.

What happened with the FTC's case against Rytr?

The FTC approved a final order against Rytr, an AI review-generation service, in December 2024 for enabling subscribers to produce false reviews. In December 2025 the FTC reopened and set aside that specific order, citing concerns about holding general-purpose AI tools liable for how subscribers might misuse them — a narrow reversal of one enforcement theory. The underlying rule banning fake reviews remains fully in effect, and businesses that publish fabricated testimonials are still liable.

Does buying followers or views count under this rule?

Yes. The rule separately bans buying, selling, or procuring fake social media indicators — followers, views, likes, reposts — used to misrepresent influence or reach, with liability attaching once there were reasonable signs the indicators were fake.

What should a compliant AI testimonial workflow actually do?

Start from a real, sourced customer quote or review and use AI to produce the video — presenter, voice, product cutaways — around those actual words. A workflow that instead asks a model to invent a customer's experience from a prompt is generating the exact fabrication the rule prohibits, no matter how the output is labeled.

If you're sitting on real reviews you've never turned into creative, that's the compliant version of this format already waiting to be made: open the AI Testimonial Video Generator and give a real quote a face and a voice.