AI in Advertising
What marketers need to know
In a short space of time, AI has joined Photoshop, CGI and traditional production as another way to make advertising.
It can put products almost anywhere. Turn stills into moving images. Create campaign variations at scale. And make ideas possible that previously needed much bigger budgets, crews or timelines.
But it also brings new choices for marketing teams.
Here’s what to consider before using AI in your production.
Where is AI actually useful?
Start by identifying where production becomes challenging.
You may need 50 assets instead of five. A campaign might need to reach various markets, with localisation. You might want the same product in ten spots. There may not be enough budget for a proper shoot. You could have existing photos but need movement. Or the idea may be too expensive to produce in the traditional way.
These are scenarios where AI shines.
Sometimes, AI can replace much of traditional production. Other times, it works better for specific tasks.
And sometimes, using a camera is still the simplest route. If you need two people talking on a sofa, filming them might still be the best option.
The burger in the ad isn't the burger in the box
A lot of the narrative around AI centres on it being ‘fake’. But advertising and marketers have always created a better version of reality.
The burger in the ad is not the burger in the box.
Food stylists prep food carefully for the camera. Stylists pin clothes behind models for a better fit. Models pose as customers. Skin is enhanced. Skies are swapped.
AI just provides another way to do this.
Marketers must still answer the same question:
How far can we enhance the presentation without misrepresenting what we sell?
The ACCC's guidance on misleading claims states that businesses must provide accurate information about their products. This includes images, descriptions and claims about qualities or performance.
Product accuracy doesn't mean every image has to be documentary truth. Advertising has never worked that way.
It means the overall impression can't materially misrepresent what you're selling.
Real or AI doesn't have to be a binary choice
A useful production model lies in the middle.
Shoot what needs to be accurate. Generate what doesn’t.
Photograph the product and create environments around it. Film the talent and change locations. Start with approved brand assets and build new formats. Use real photography for the main image and generate variations around it.
This approach combines the benefits of traditional production with the flexibility of generative techniques.
There’s no need to generate every pixel.
What about AI-generated people?
Advertising has never required every person in an ad to be a real customer.
We hire actors, cast models and fill spaces with extras. Someone pretending to enjoy an insurance renewal on camera didn’t just wander into the shot.
So, using a synthetic person doesn’t automatically make an ad deceptive.
A generated woman drinking a soft drink isn't inherently misleading.
But a generated woman claiming the drink cured her migraines is a different story.
The problem isn't that the person is synthetic. The problem begins when the audience is led to believe the person's experience, identity or endorsement is real.
That’s not a new AI issue. It’s about advertising claims and testimonials, which are now easier to produce.
The same Australian Consumer Law principles apply to misleading claims. The ACCC includes testimonials, social media and images in its accuracy requirements.
Synthetic people need more attention when audiences could reasonably believe their experiences or endorsements are real.
For clearly fictional content, the calculations change.
Can you actually keep AI on brand?
For one image, yes.
At 500 images, things get trickier.
Generative systems excel at creating variations. However, brands usually don’t want every variable to change.
Colours can drift. Faces may vary. Products might mutate. The lighting could shift to another campaign. A polished image can turn glossy and artificial after several generations.
AI production often turns into an art direction challenge, not just a prompting one.
Decide what can change and what must stay the same.
This could mean fixing colours, lighting, composition, product treatment and other visual rules, while allowing flexibility elsewhere.
In a recent project, Nothing Is Real created a system built for over 1,000 images. The solution wasn’t 1,000 perfected prompts. We first established a visual world with rules around colour and lighting, then added flexibility within those limits.
The more content you create, the more important those limits become.
Research published by Canva in 2026 found that 94% of marketers were concerned about brand consistency as AI use scales, while only 9% trusted their tools to handle it well.
Will customers care?
Many say they do.
But how the research is worded is important.
A study by Ad Standards and Roy Morgan published on 6 May 2026 found that 72% of Australians were worried about AI-generated content in ads. They were especially concerned about it misleading people with fake or unrealistic content.
A separate YouGov study published on 20 May 2026 found that 45% of Australians would trust a brand less if its advertising was mainly created using AI. It also revealed that 40% were uncomfortable with brands primarily using AI for ads, while 25% felt comfortable.
“Mainly using AI” is crucial.
Neither result suggests that consumers object to every AI-assisted retouch, generated background or production shortcut.
They indicate that marketers should not assume audiences are indifferent.
The same YouGov study found that 86% of Australians believed brands should clearly disclose when their advertising was mainly created using AI.
There’s a big difference between finding out a fantastical campaign world was generated and discovering that a supposed real customer never existed.
Do you have to tell people?
There’s no universal rule that every use of AI in an advertisement needs a giant AI sticker.
Australia's National AI Centre recommends choosing the level of transparency according to the context, including how content was created or altered and its potential impact on the audience.
Higher-impact content or AI changes that affect meaning may require more transparency. Low-impact content with limited AI involvement may need little or no disclosure.
Industry guidance is moving in a similar direction.
The IAB AI Transparency & Disclosure Framework V2 takes a risk-based approach rather than recommending labels for every AI use. It also warns against unnecessary labelling, which can lead to disclosure fatigue.
Platforms have their own systems too.
On 9 July 2026, Google announced new AI transparency for ads, including a “How this ad was made” section in My Ad Center across Search, YouTube and Discover.
On 1 June 2026, Meta announced an expansion of its advertising transparency system. Meta's “About this ad” system shares details when its generative tools have played a significant role in creating or editing an ad. Meta is also adding signals from third-party AI tools.
The details change quickly, so we've made a regularly updated guide:
IS YOUR AI CONTENT BEING FLAGGED? →
What about copyright?
Copyright is one of the less settled areas of AI production, so be wary of simple answers.
For marketing teams, the important question is usually practical: can we safely use this work commercially?
If you're commissioning AI-assisted work, ask the same kinds of questions you would of any production partner. Where did the key assets come from? Are there third-party images, characters, logos or other protected material in the final work? Does the production team have the necessary commercial rights to the tools and assets they've used?
It’s also worth knowing that permission to use an AI tool commercially and copyright ownership of its outputs are not necessarily the same thing.
For everyday content, this may be relatively straightforward. For a major campaign, a valuable brand asset or work you intend to own and reuse for years, the stakes are higher.
The bigger the investment and the more important ownership is, the more carefully those questions should be checked.
Where should a marketing team start?
Don’t begin with a list of AI tools.
Examine how you already create content.
Where does production become too expensive? What takes too long? What content never gets made? Which ideas get diluted because of logistics? Where are you producing the same asset again and again?
Then identify what genuinely needs to stay fixed.
The product? A specific person? An actual location? Exact brand assets? Nothing at all?
The useful question isn’t “Should we use AI?”
It’s “Where could AI make this better, faster or more achievable without compromising what needs to remain true?”
Sometimes the answer is almost everywhere. Sometimes it’s one background. Sometimes a camera is still the better tool.
Knowing the difference is the important part.
Want to see what AI could do for your next project?