What Are The FTC AI Content Rules In 2026?

In January 2026 the FTC established a dedicated AI enforcement unit. AI-generated testimonials are prohibited outright.

The FTC did not ban AI. It demanded transparency.

This is the part that gets misreported constantly, so let me be precise.

The FTC's position is not that AI tools are prohibited. It is that if AI produced or substantially modified the content a consumer reads, the consumer should know.

There is no AI exemption from the existing rules. Deception is deception whether a person or a model wrote it, and the disclosure obligation follows the material connection rather than the author.

In January 2026 the FTC established a dedicated AI enforcement unit. This is not theoretical.

I am not a lawyer and this is not legal advice. What follows is published FTC guidance and documented penalties, which are matters of public record. How they apply to your specific business is a question for somebody qualified to answer it.

Double disclosure, explained properly

When AI is involved in sponsored content, you may need **two separate disclosures**. The FTC calls this layered disclosure. The industry calls it double disclosure.

**Disclosure one: the material connection.** The traditional one, required since the original Endorsement Guides. If there is payment, a free product, an affiliate commission, or an employment relationship, it must be disclosed. Standard formats: #ad, #sponsored, 'Paid Partnership', or a plain statement.

**Disclosure two: the AI involvement.** If AI generated or substantially modified the content, that should be disclosed too.

Neither substitutes for the other. A post that discloses the commission but hides that a model wrote the review has disclosed one of two things it needed to.

Where the AI disclosure has to sit

The guidance on this is stricter than most people assume, and it maps onto the same four-factor test.

'Visible' means near the top of a review, in the opening seconds of a video, or prominently in a social caption.

**A footer mention is not sufficient** under the reasonable consumer standard. Neither is a hashtag buried among fifteen others.

The test is whether a normal person, reading normally, would know before they reached the recommendation.

The line you cannot cross: AI-generated testimonials

**AI-generated testimonials are prohibited outright. Disclosure does not make them compliant.**

The FTC treats them as fake reviews. The reasoning is straightforward: a testimonial is supposed to reflect the honest opinion of a real person who actually used the product. An AI-written one does not, however clearly you label it.

Adding 'generated by AI for illustrative purposes' does not fix it. The problem is not the disclosure. The problem is that the opinion does not exist.

The same applies to synthetic influencers and cloned voices. An undisclosed virtual endorser is considered inherently deceptive, because audiences cannot evaluate the credibility of somebody who is not real.

So the rule for your business is simple and absolute. Use AI to help you write. Never use AI to invent a customer.

What needs disclosing and what does not

The test is whether the AI assistance materially affects how a consumer interprets the content.

**Cosmetic changes do not require disclosure.** Grammar correction. Formatting. Tightening a sentence you wrote.

**Substantive generation does.** If the model produced the argument, the review, or the recommendation and you lightly edited it, that is substantive.

The honest question to ask yourself is this: if a reader knew how this was made, would they weigh it differently? If yes, disclose.

There is also a state-level layer now. New York passed a dedicated AI content disclosure law for advertising, signed in January 2026 and effective June 2026. State laws apply to content consumed by residents of that state, which for a national audience means complying with the strictest applicable standard.

What I do on this site

I use AI to help me write. I do not use it to invent an opinion, a customer, or a result.

Every review here carries an affiliate disclosure at the top, before the recommendation, in plain language. Every review states plainly where the product I earn nothing from is the better choice.

And there are no case studies on this site that were not built from real numbers, because inventing one would be fabricating evidence — precisely the thing this article is about.

The reason is not fear of a penalty. It is that an audience which suspects you are for sale stops being worth anything, and that loss does not have a payment plan.

Why most owners never do this

The obstacle is rarely knowledge. Most people running this broadly understand what they should be doing, and the gap between knowing and doing is where the difficulty sits.

Part of it is that nothing forces you to. Nobody sends a reminder about your disclosure or your business profile. So it waits, and it can wait for years.

Thirty-three percent of my audience told me they research all day and never start. Research feels like progress and carries no risk of being wrong in public.

So make it deliberate. Put a specific hour aside. One thing done properly beats a list you never start.

The opening month, mapped out

Week one, work out the number. Whatever this is about, calculate your own version rather than accepting somebody else's.

Week two, do the free thing. There is almost always one and it is usually the highest-return hour available.

Week three, put something real in front of somebody real. An offer, with a price and a way to pay.

Week four, look honestly at what happened and fix the one thing that clearly went wrong.

That is a month, it costs nothing, and it moves you further than another year of reading would.

The timeline, so you can see where this is heading

The core principle has not changed since 1980. A material connection between somebody endorsing a product and the business behind it must be disclosed, so consumers can weigh the endorsement knowing what sits behind it.

What has changed is scope and enforcement. The Endorsement Guides were revised in June 2023, broadening the definition of endorsement and material connection in ways that capture content most small operators would never have thought of as advertising at all.

The Consumer Review Rule took effect in late 2024, addressing fake reviews, incentivised testimonials, and the suppression of negative feedback. In March 2025 the FTC published staff guidance specifically on the use of AI in advertising.

In December 2025 it sent warning letters to ten companies, naming four problems: fake reviews, undisclosed incentives for positive posts, insider testimonials without clear labels, and active suppression of negative feedback.

In January 2026 it established a dedicated AI enforcement unit, and the maximum penalty rose to $53,088 per violation. None of that is speculation. It is the published record, and the direction is unambiguous.

Why the honest version converts better anyway

Here is the part that gets lost in every compliance conversation. Being transparent about a paid relationship does not weaken the recommendation. It strengthens it.

A reader who can see that you earn from a link, and who then watches you say plainly that a competitor is the better choice for their situation, learns something important about you. They learn that you can be trusted, and that trust is what makes the next recommendation land.

The people who hide the relationship are protecting a short-term conversion rate at the cost of the only asset that genuinely compounds. And readers are considerably better at detecting it than most publishers assume.

So do it properly because it works, not merely because it is required. The compliance is a by-product of the honesty rather than the reason for it, and that is a considerably more comfortable way to run a business.

What to check on your own site this week

Look at it with fresh eyes. Where is the disclosure? Is it above the recommendation or below it? Could a normal person, reading normally, miss it entirely?

Read your testimonials page honestly. Who wrote those? Did anybody receive anything in exchange, including a free product or a discount?

Check whether you have ever removed a negative review because it was negative. That is suppression, and it was one of the four things the FTC named.

And if AI writes content that recommends things, decide now how you will handle that. None of this costs money. All of it is worth an afternoon.

Questions I get asked about this

How long before this starts working?

Longer than most people hope and sooner than most people fear. Give it a few months of steady effort rather than a few weeks, and judge it on whether enquiries are trending up rather than on any single week.

Do I need to spend money to start?

No. The highest-leverage steps here are free, and they are the ones almost nobody does. Tools help you measure and move faster. They do not do the work.

What if I have no time?

Then pick one thing from this and do it properly rather than five things badly. One change that actually happens beats a list you never start.

Is AI going to replace this work?

Not the work itself. It is genuinely useful for the repetitive parts around the work, and that is where the hours come back. The judgment stays yours, and that is what you are being paid for.

Affiliate Disclosure: This article uses affiliate links and may earn a commission from certain links, at no extra cost to you. Opinions expressed are our own.