AI-Generated Images in Your Adverts: What the ASA Expects
AI image tools have made it easy for a small business to produce a polished advert in minutes: a dreamy shot of a candle on a mantelpiece, a smiling customer holding a product, a before-and-after for a treatment. The Advertising Standards Authority's position is simple. The UK advertising code applies however an advert was made. In a June 2026 piece for advertisers, the ASA and its sister body CAP said the code is media-neutral, and that advertisers remain responsible for the final result even when an automated tool produced it. So the useful question is not "did we use AI?" but "would this image give a customer a false impression?"
What the regulator has said
The June 2026 guidance picks out four points worth knowing:
- The rules are the same. Misleading advertising, substantiation and social responsibility rules apply to AI-made images just as they do to photographs or illustrations.
- Fake endorsements are a problem. A synthetic version of a well-known person appearing to recommend your product is likely to breach the code, and using someone's likeness without permission raises legal issues beyond the ASA's remit.
- Bias in outputs is your problem. Image tools can produce stereotypes. Checking outputs for harmful or offensive content is part of compliance.
- Automation does not shift responsibility. If a platform generates or places the ad, you are still the advertiser.
The guidance does not set out a general rule that every AI image must be labelled. The test is whether the overall ad misleads.
A recent ruling worth reading
In September 2026 the ASA upheld complaints against two paid social ads for an appearance app that used before-and-after faces with scores. The ASA had seen no evidence that the images were genuine, that the before and after showed the same person, or that they reflected typical results, and found the ads misleading and exaggerated. The ruling turns on substantiation rather than on how the images were made. That is exactly the trap with generated images: an AI tool will cheerfully produce a perfect "after" picture, but you cannot hold evidence that it shows what your product does.
Where small businesses are most exposed
Product appearance
A generated image of your product that looks better than the real thing is the most common risk. If a cushion's fabric appears richer, a cake taller or a room bigger than reality, the image is making a claim. Customers who receive something different will feel misled, and that is the core of a complaint.
Results and before-and-after
Any image that shows what a service achieves, such as a cleaned driveway, a renovated kitchen or a skin treatment, needs to reflect real, typical results. Use your own photographs of real jobs, with the customer's permission, and keep the originals.
People who do not exist
A generated "happy customer" holding your product can imply a testimonial that never happened. A generated "team photo" can suggest a business is larger than it is. Neither is automatically a breach, but both need thought about the impression they create.
Mood and scene-setting
The lowest risk is clearly decorative imagery: a winter landscape behind a seasonal offer, an abstract pattern, a stylised illustration. Nobody reads a painted-looking snowy village as a claim about your product.
A check before anything goes live
Run each AI-assisted ad through these questions before it is posted or printed:
- Does the image show our actual product, service or result? If not, is that obvious to a reasonable viewer?
- Could someone buy based on a detail in the image that is not true, such as size, colour, finish or quantity?
- Does it show a person who appears to be a customer, expert or well-known figure? Do we have a real basis for that?
- If it shows results, do we hold evidence that those results are genuine and typical?
- Would we be comfortable if a customer put our image next to what they received?
- Has a person, not just the tool, looked at the final version at full size, including hands, text in the scene and background details?
Keep a note of the prompt, the tool and the date for each image you use in advertising, alongside any evidence for claims. If a question ever arises, being able to show how the image was made and why you thought it fair is useful.
When a label helps
Even without a blanket labelling duty, a short note such as "illustration" or "image for illustration only" can stop a scene-setting image being read as a product photo. It will not rescue an ad that is misleading overall, which is the point the ASA keeps returning to. A disclaimer cannot contradict the main impression an image creates.
Keeping your brand honest and recognisable
There is a brand argument too. Generated imagery tends to look generic, and a feed full of it can make a small firm look like everyone else. Real photographs of your premises, your people and your work carry a credibility that is hard to fake. If you use AI tools, use them for the parts where they help, such as backgrounds, layout ideas and mock-ups, and keep the product itself real. Our article on building a recognisable brand on a small budget covers how to keep a consistent look across ads, and why print still matters applies the same thinking to flyers and posters, where the code applies just as it does online.
If you are planning a larger campaign or working in a sensitive area such as health, beauty or finance, a specialist advertising adviser can look at specific ads before they run.