When an AI agent writes your ads, drafts your reviews page or publishes your SEO content, a reasonable question follows: who is responsible if the output misleads someone? Regulators in the US and UK have addressed this in their own words. Neither the FTC nor the UK Advertising Standards Authority treats AI as a reason to apply different rules, so the advertiser stays accountable for what the agent produces.
- The FTC’s August 14, 2024 final rule on fake reviews bans fake or false reviews and testimonials, including AI-generated ones.
- The FTC’s September 25, 2024 Operation AI Comply announced five actions, including a proposed $193,000 settlement with DoNotPay over unsupported “robot lawyer” claims.
- The UK ASA said in February 2025 that its Codes already cover AI-related issues such as misleading claims, misleading endorsements and offensive imagery.
- The practical control is process: substantiate claims, restrict what agents may publish, and keep a human sign-off on high-risk content.
Scope note: this is a marketing-operations summary of public regulator statements, not legal advice. The FTC items date from 2024 and the ASA item from February 2025, and we did not review later enforcement. Check current guidance or ask counsel before relying on any of it.
What the FTC has said
Two FTC announcements matter most to marketers using AI. The first is the final rule on fake reviews and testimonials, announced on August 14, 2024. The FTC’s release says the rule bans fake or false reviews, including AI-generated ones, and covers creating, selling, buying or spreading them when the business knew or should have known they were fake. It also lets the FTC seek civil penalties against knowing violators.
The second is Operation AI Comply, announced September 25, 2024, which the FTC described as five actions against deceptive AI claims and schemes. One involved DoNotPay, which the FTC alleged had marketed its service as the world’s first robot lawyer without evidence it could replace human lawyers. The proposed settlement required a $193,000 payment.
Then-Chair Lina Khan said at the time that using AI tools to trick, mislead or defraud people is illegal. For a buyer of agentic services, the implication is direct: an agent that generates testimonials or overstates what a product does creates exposure for the business that publishes the output.
What the UK ASA has said
The UK Advertising Standards Authority, which enforces the CAP and BCAP advertising codes, published an article on AI in advertising on February 7, 2025. It says existing Codes already cover many AI-related issues, including misleading images and claims, misleading endorsements, and harmful or offensive imagery. AI-made images are judged by the same standards as any other image.
The article adds that rules on depicting living people would likely apply to AI depictions of recognizable individuals, although the ASA had not yet ruled on that point. It also describes the ASA as a proactive regulator that monitors emerging issues and runs a Scam Ad Alert System with major online platforms.
The article gives no enforcement statistics, so we cannot say how often AI-related complaints are upheld. What it does establish is the principle. Where a UK ad would be misleading if a human wrote it, it is equally misleading when an agent writes it, and the advertiser answers for it.
US and UK side by side
The table below compares what each regulator’s cited statement establishes for a marketer. It reflects only the sources named above, so it is a starting map and not a full compliance matrix.
| Risk area | US (FTC statements cited) | UK (ASA article cited) |
|---|---|---|
| Fake or misleading endorsements | Final rule bans fake or false reviews, including AI-generated ones | Existing Codes cover misleading endorsements |
| Overstated AI capability | DoNotPay action over unsupported “robot lawyer” claims | Not addressed in the cited article |
| Misleading claims and images | Deceptive AI claims challenged under Operation AI Comply | Same standards apply to AI-made images and claims |
| Who is accountable | The business making or publishing the claim | The advertiser, under media-neutral Codes |
Controls to set before an agent publishes
None of the sources above prescribes a workflow, so the steps below are our recommendation based on them. They aim at the two failure points the regulators describe: unsupported claims and fabricated endorsements. Each can be assigned to a person and checked within a week.
- Keep a claims register listing every product or performance claim an agent may use, with the evidence behind each.
- Block agents from generating reviews, testimonials or case-study quotes unless a real, attributable customer supplied them.
- Require human sign-off for new claims, comparative statements and any content depicting a real person.
- Log what each agent published and when, so you can review or retract it quickly.
- Ask vendors in writing who is responsible for substantiation and how they handle takedowns.
Our article on AI-generated ad creative covers performance evidence, and our guide to agentic SEO and Google’s spam policies covers the platform-policy side of the same problem.
Frequently Asked Questions
Are AI-generated reviews illegal in the US?
The FTC’s final rule, announced August 14, 2024, bans fake or false reviews and testimonials, including AI-generated ones, when the business knew or should have known they were fake.
Does the UK have separate advertising rules for AI?
The ASA stated in February 2025 that its existing Codes already cover many AI-related issues. AI-made images and claims are judged by the same standards as any other advertising.
Who is responsible when an AI agent publishes a misleading claim?
The regulator statements cited here place responsibility on the advertiser or business making the claim. Contracts with vendors can allocate costs, but they do not remove that accountability.
Handing work to an AI agent changes who types the words, not who answers for them. At Grow with Hafsa we help US and UK teams build claim registers, approval steps and publishing logs into their AI marketing and agent workflows, so automation does not outrun review. A useful companion read is our guide to evaluating an agentic marketing vendor.
The same accountability applies to reviews: see where the FTC and UK fake-review rules draw the line for GEO.
Related: for a contract-side view of the same question, read Meta’s AI Features terms: what is confirmed and what is on hold.