AI can help humans in many ways, such as breakthroughs in drug discovery and digital voice assistants, but it has its limits. An AI assistant cannot actually “offer a helping hand” or “lend an ear”, but it will try to reply with the best response as predicted by its programming.
Many people are familiar with the Three Laws from The Handbook of Robotics, 56th Edition, but it is not the ASA’s role to come up with laws for AI assistants. Unless and until there is AI-specific legislation, how can we apply the CAP Code as it currently exists to ads for AI assistants?
Foundational knowledge
A fundamental principle is that the CAP Code only applies to ads, not the products advertised. Humans may find a product or service objectionable in itself, but the ASA has ruled that legitimate businesses are permitted to advertise, provided they do so responsibly.
This means that we cannot comment on the (in)actions of the AI product itself. The ASA is limited to assessing how AI products are presented in ads.
The Troubles of Modern Life
AI assistants sometimes have a human-like avatar for the user to interface with, and sometimes they appear more like an instant messenger programme. In either case, the AI assistant is a software application that takes input from the user and gives a real-time response, i.e. a chatbot. These products are often standalone apps, or features within an app with additional functions.
There are a few broad use categories when it comes to AI personal assistants, which can overlap:
- Productivity – designed to automate mundane tasks.
- Companionship – simulating a friend, or romantic partner.
- Health and wellness – e.g. virtual dieticians, life coaches, therapists.
“Hi, what can I help you with?”
Some AI products claim to offer solutions to the more mundane or frustrating admin aspects of a human’s job. With something as nebulous as “productivity”, the ASA may find that the ads are putting forward obvious exaggerations (AKA puffery) or clearly subjective claims, which may not be an issue under the CAP Code. For example, a car insurance ad claimed that this insurance was so good that it “really solves your problems”, which the ASA found to be puffery.
There are also AI products that claim that they can generate income with minimal human effort, such as programmes that will play the stock market for you, or generate art that it can sell for you. While a novel application, the ASA has seen similar promises before – unrealistic earnings claims have been ruled misleading when made by ads for multi-level marketing companies, and courses aimed at business owners to improve their revenue.
Essentially, what the ad promises, the product should deliver.
Can I Copy/Paste Your Homework?
There have also been many sightings of children using AI to do their homework for them. Essay mills were made illegal in England in 2022. The ASA has upheld complaints against ads claiming that students could submit essays they purchased as their own and with no risk of plagiarism. If the ASA determined that an ad for an AI “homework helper” showed something closer to a robotic ghost writer than a spellchecker, they might consider the ad to break the Code.
Ensure that your ad does not offer students an easy A(I), as this could be seen as irresponsible.
Science Fiction Lover
The ASA has ruled against ads for apps where users can create AI characters to engage in sexually suggestive chats, or simulated videos of the user’s crush, because these ads objectified women.
Whether an AI can offer companionship may be a question better left to philosophers. However, there are AI chatbots that claim to offer dating advice. Such advice comes in many forms, and can range from uncontroversial suggestions for self-improvement, to dehumanisation and violence. It stands to reason that ads for AI dating advice could invoke harmful gender stereotypes, or otherwise problematic content, especially when AI is programmed by humans with internal biases.
For example, an ad might not only use an avatar objectifying women to promote their AI dating assistant but could perpetuate harmful stereotypes about men and women. Ofcom released a research report into the manosphere in 2025, which indicated that a lot of men became involved in the manosphere while seeking relationship advice. An ad that suggested that a man was inferior because he needed dating advice could be interpreted as mocking a man for not conforming to a gender stereotype, which is a guiding principle of what to avoid from CAP’s Advertising Guidance on depicting gender stereotypes. The ASA has employed this principle when an ad was found to have belittled a woman running a business, so they would be unlikely to accept an ad that perpetuated a harmful stereotype against men.
Make sure your ad is not some conjurer of cheap clicks. An ad for an AI product offering dating advice should not degrade by design.
If Looks(max) Could Kill
Ofcom’s report also discussed “looksmaxxing”: men trying to change their appearance to meet an idealised standard, informed by other participants in the manosphere. Looksmaxxing communities might share recommendations for diet and exercise, skincare, and sometimes plastic surgery. Before AI assistants, men would upload their photos in these communities to receive tailored (but not necessarily constructive) feedback.
An AI assistant could take that photo and analyse a face down to the millimetre. If an ad claimed the AI could give similar advice, this could be considered exploiting insecurities, or creating pressure to confirm to idealised, gendered physical features. The ASA has upheld against ads that have exploited body image insecurities previously. Even if an ad is not for a cosmetic intervention itself, it must still be socially responsible.
Do no body harm - ensure your ads do not perpetuate harmful body image.
Pla-C3-b-0
The ASA has already upheld against an ad for AI product that made medical claims without the required authorisation, and discouraged essential treatment for a medical condition.
AI chatbots have been proposed as a potentially useful tool for people waiting to talk to mental healthcare professionals. While the evidence is still being gathered, we would recommend avoiding ads that imply that an AI service can replace a qualified therapist. We would strongly discourage claims that these apps would be suitable for emergency care or crises.
When it comes to more general wellness, the ASA has previously ruled that ads for health plans which promised unrealistic outcomes were misleading. Claims that a specific amount of weight loss can be achieved in a given time period are already not permitted.
Unless you can provide evidence of the product’s health hacks, keep your AI from playing doctor.
Scanning is Believing
Ads may show things that AI assistants can do, which advertisers would have to substantiate. Many AI assistants claim that they can extract and summarise information from a photo, but accuracy may be variable.
For example, an AI diet assistant might claim it could calculate a nutritional breakdown from a photo of a plate of food. If the advertiser could not present evidence of the AI’s accuracy, the ad could be found to be misleading, or even irresponsible if it led to poor nutrition. The ASA has previously found ads that proposed replacing all food with meal replacement drinks was irresponsible, because the products would not provide sufficient calories if they replaced all meals.
Furthermore, an app might scan and analyse the photo for free, but only offer the advice after payment. If the ad did not make clear that the full service would only be provided after payment, the ASA would be likely to conclude that the ad was misleading for omitting material information.
Calories count for humans, so make sure your ads only dish out facts.
Human Advice vs Machine Learning
A human summary of the advice for marketers of AI products when preparing their ads:
- Hold evidence to back up any objective claims about what your product can do; and
- Make sure the content is responsible.
Even shorter: Keep it real in your ads, even when advertising AI products.
Hopefully you’re now feeling more confident about where the line is drawn. However, if you notice any uncertainty starting to hallucinate around the edges, the Copy Advice team can help re‑anchor your ads.
More on
-
Keep up to date
Sign up to our rulings, newsletters and emargoed access for Press. Subscribe now.
