AI in marketing and sales: honest ads, real reviews and lawful outreach

Cue 1 An AI generated advert passing through the same four tests as any other: legal, decent, honest, truthful

EMMA 00:00 George, I generate half our ad images with AI now. Are there rules about that?

GEORGE 00:06 No AI rules, and that's the point. The Advertising Standards Authority enforces the CAP Code, which covers your website and social media too. It has no AI specific rules, because the existing ones apply however the content was made.

EMMA 00:21 So what applies?

GEORGE 00:22 Everything that always did. Legal, decent, honest and truthful. Don't materially mislead. Be able to substantiate your claims. The same rules in the same way.

Cue 2 Two cases side by side: an AI influencer that needs disclosing and a true sentence that does not

EMMA 00:32 Do I have to label AI images?

GEORGE 00:35 There's no blanket rule. The ASA suggests a question. Is the audience likely to be misled if the use of AI isn't disclosed?

EMMA 00:44 Give me a yes and a no.

GEORGE 00:47 A yes. Your influencer is AI generated and people think she's real. Saying so removes a misleading impression. A no. AI helped write a sentence of copy that's true. Nobody's misled.

Cue 3 A cosmetic before and after image with a made with AI label, still marked misleading

EMMA 00:59 So if I'm worried, I just add a label.

GEORGE 01:03 That's the mistake. The ASA says disclosure alone is very unlikely to fix a fundamentally misleading message. You can't make a misleading claim and then disclaim it by saying AI was used.

EMMA 01:16 Example?

GEORGE 01:16 Their own. A cosmetic product. An AI image showing an effect that doesn't reflect real results is likely to be materially misleading. A line saying made with AI doesn't rescue it.

Cue 4 A real person's face on an advert with no permission, and an invented customer quote

EMMA 01:29 We sell holiday lets. I've been brightening the rooms.

GEORGE 01:32 Brightening is one thing. A room that's larger than the real one is another. If the picture shows what the product is, it has to be what the product is. A label doesn't change that.

EMMA 01:46 What about putting a celebrity's face on it?

GEORGE 01:50 A real person's face or voice, without permission, suggesting they endorse you. That breaks the rules on endorsements and may breach other law too. And an invented customer with an invented quote is a fake testimonial, however lifelike.

Cue 5 An advertiser, an agency and a tool, with the responsibility arrow pointing at the advertiser

EMMA 02:05 If my agency makes it, is it their problem?

GEORGE 02:08 Yours. The advertiser is responsible. It doesn't matter that an agency made it, a tool generated it, or a platform assembled the wording automatically. If it appears as your marketing, the ASA asks you to substantiate it.

EMMA 02:23 Give me one check I can do every time.

GEORGE 02:27 Ask one person who didn't make it. What does this tell you about the product? Then check that what they say is true.

Cue 1 Fifty generated five star reviews crossed out, with the date 6 April 2025

EMMA 02:36 George, a supplier offered to generate fifty five star reviews for our new product. Tempting. Legal?

GEORGE 02:42 No. Since 6 April 2025, under the Digital Markets, Competition and Consumers Act, it's a banned practice to submit or commission fake consumer reviews. In all circumstances.

EMMA 02:53 What counts as fake?

GEORGE 02:55 A review written by a tool for a customer who doesn't exist, or who never said it. And a real review rewritten to be more favourable than what the customer wrote.

Cue 2 A review page hiding its one star reviews, and an AI summary leaving them out

EMMA 03:07 What about incentives? We give a discount for a review.

GEORGE 03:11 Allowed, but it must be clearly identified as incentivised. And you can't publish reviews in a misleading way.

EMMA 03:18 Meaning?

GEORGE 03:19 Hiding the bad ones. Or an AI summary of your reviews that leaves them out. That presents review information in a misleading way.

Cue 3 A business checking that a reviewer was a customer, with a report button

EMMA 03:28 And the penalty?

GEORGE 03:29 The Competition and Markets Authority can now decide for itself that consumer law's been broken, and fine a business up to ten per cent of its global annual turnover. Without going to court.

EMMA 03:42 That's serious. Is not writing them enough?

GEORGE 03:45 No. If you publish reviews, you're expected to take reasonable and proportionate steps to prevent fake ones and remove them when found. A policy. A way to check a reviewer was a customer. A way to report a suspect review.

Cue 4 A bought list being personalised by a tool, stopped at a consent gate

EMMA 04:01 Different subject. I've got a tool that writes a personalised email to every lead on a list I bought.

GEORGE 04:09 AI can write the message. It can't give you permission to send it. Email and text marketing to individuals needs their consent, with a narrow exception for your own existing customers.

EMMA 04:21 But it's personalised.

GEORGE 04:22 Which makes it worse, not better. Personalisation uses personal data. And a bought list personalised by a tool is still an unlawful campaign.

Cue 5 A chatbot with a human name and photo, relabelled clearly as an AI assistant with a route to a person

EMMA 04:32 Who enforces that?

GEORGE 04:33 The Information Commissioner's Office. And the Data (Use and Access) Act 2025 raised the maximum fines to the same level as data protection law.

EMMA 04:42 Last one. Our website chatbot has a name and a photo. People think it's a person.

GEORGE 04:49 Then fix that. Presenting an automated system as a human being is misleading. And if you serve customers in the EU, the AI Act has required since 2 August 2026 that people are told they're dealing with an AI system.

EMMA 05:05 And if it gets a refund policy wrong?

GEORGE 05:08 Then you told the customer that. A chatbot speaks for the business. Test what it says, keep a record, and give people an easy route to a human being.