We’ve started seeing a pattern on sites that come to us for a redesign: a hero image of a smiling tradie who doesn’t exist, standing in front of a house that isn’t in Australia, holding a tool that doesn’t quite work. The owner made it in five minutes with an AI image generator because they didn’t have a decent photo. It’s understandable. It’s also usually a mistake.
AI images have a place on business websites. But there are real questions about ownership, misleading customers and trust that are worth thinking through before you use them. This is general information, not legal advice, and the law in this area is still moving.
Do you own an AI-generated image?
Possibly not, at least not in the way you own a photo you took.
Australian copyright law protects original works that have a human author. In the Telstra v Phone Directories case in 2010, the Federal Court found no copyright in directories that had been largely produced by a computer process rather than by people. How that principle applies to images made with an AI prompt hasn’t, as far as we’re aware at the time of writing, been settled by an Australian court. In the United States, the Copyright Office has refused to register works generated by AI without meaningful human authorship.
The practical upshot: if an image is purely AI-generated, you may have little or no copyright to enforce. A competitor could use the same image, or something very similar, and you might not be able to stop them.
That matters most for anything that’s meant to be distinctively yours. Never use a purely AI-generated image as your logo or the core of your brand. You want clear ownership of those, and a designer who can hand over the original files and the rights.
Could an AI image infringe someone else’s rights?
This is the murkier side. Many image generators were trained on huge collections of images scraped from the web, and there are lawsuits overseas over whether that training was lawful. How those cases end will shape the rules for everyone.
You can reduce your own risk:
- Read the tool’s terms. Check that commercial use is allowed on the plan you’re using, and whether the provider offers any indemnity. Some, such as Adobe Firefly, are marketed as trained on licensed content specifically for commercial use.
- Don’t prompt for a living artist’s style, a brand or a real person. “In the style of [artist]” or “looks like [celebrity]” is asking for trouble.
- Watch for logos and text. Generators sometimes produce something close to a real brand’s logo or packaging in the background. Look closely before you publish.
Could an AI image mislead customers?
This is the risk we’d take most seriously, because it’s covered by laws that already exist.
Under Australian Consumer Law, businesses must not engage in misleading or deceptive conduct, and that includes images. The ACCC’s guidance on false or misleading claims is about the overall impression your advertising creates, not just the words.
Some examples of where an AI image could cross that line:
- A “recent project” gallery for a builder that includes generated images of homes they never built.
- A team photo showing more staff than the business has.
- A cafe menu with generated food photos that look nothing like what arrives at the table.
- A product image showing features, finishes or colours the product doesn’t have.
- An accommodation listing with a generated view that isn’t the real view.
For health practitioners the bar is higher again. Advertising of regulated health services is overseen by Ahpra and the National Boards, and images that create unreasonable expectations of results are a problem. A generated “after” photo, or a generated image of a happy patient implying an outcome, is exactly the kind of thing to avoid.
A simple test: if a customer found out the image was AI-generated, would they feel misled about what they were buying? If yes, don’t use it.
Do you have to disclose AI images?
At the time of writing, there’s no general Australian law that requires a business to label every AI image on its website. The misleading conduct rules above are what matter: if the image could give a false impression, a disclaimer in small print probably won’t save it, and you shouldn’t use it.
Platforms have their own rules. Meta applies “AI info” labels to some AI-generated or edited content on Facebook and Instagram, and may detect it automatically. If you post the same image on social media, expect it might be labelled there even if it isn’t on your site.
Our view: where an image is clearly illustrative (a concept graphic, a stylised background), disclosure isn’t needed. Where it could reasonably be taken as a real photo of your work, people, premises or products, either disclose it or, better, replace it with a real photo.
The practical problems
Even when the legal side is fine, AI images often let a site down in ways owners don’t notice.
They look wrong for Australia. American-style houses, power points with the wrong pins, left-hand-drive utes, snow-dusted trees in a Ballina summer, cars on the wrong side of the road. Local customers notice, even if they can’t say exactly what’s off.
The details give them away. Hands, tools, writing on signs and vans, the number of legs on a chair. A tradie holding a drill that couldn’t possibly work tells customers you didn’t have a real photo.
They’re the same as everyone else’s. Generators lean towards a recognisable look: glossy, oversaturated, perfectly lit. Customers increasingly read that look as “not real”.
They can slow your site. Generated images often export as large PNG files. Resize them and convert to WebP before uploading, as we explain in why your website is slow and what actually fixes it.
Where AI images are fine, and where to avoid them
| Use | Our view |
|---|---|
| Abstract backgrounds, textures, patterns | Fine |
| Illustrations for blog posts about general topics | Fine, if clearly illustrative |
| Concept images (“what a modern laundry could look like”) | Fine with a label |
| Social graphics with no people or products | Generally fine |
| Your logo or brand mark | Avoid. Ownership is unclear |
| Photos of your team | Avoid. Use real people |
| Project galleries and before-and-afters | Never |
| Product photos | Never as the main image |
| Health outcomes or patient images | Never |
Why real photos still win
People hire people. A real photo of the owner, the van, the shopfront or the finished job tells a customer this business exists and does what it says. For a contractor like Field Directional Drilling, the real rigs and real sites are the proof of capability; nothing generated could stand in for them.
Real photos don’t need to be expensive. A phone photo taken in good light, straight, with the clutter moved out of shot, beats a perfect AI image for trust every time. Professional photography is worth it for the hero images and team shots that set the tone of the whole site, which is why we keep photographers in-house for commercial shoots and team headshots.
If you’re using AI for writing as well, the same principle applies: our guide to using ChatGPT for marketing without sounding like a robot covers keeping it honest.
Before your next site update, go through your current images and sort them into three piles: real, clearly illustrative, and “could be taken as real but isn’t”. The third pile is the one to replace. Planning which shots a new site needs, and when to book the photographer, is part of every web design project we run. If your third pile is a big one, we can talk it through.