Can I Use AI-Generated Images in Advertising in Australia?
Short answer: yes, you can use AI-generated images in Australian advertising, but they have to clear the same legal bar as any other creative. There is no AI-specific ad law in Australia yet, so the rules that already apply, the ACL, AANA Code of Ethics, Copyright Act and Privacy Act, still govern every frame you publish. Absolutely AI runs client work against a pre-flight checklist to keep AI imagery on the right side of all four.
Australian marketers keep asking the same question in slightly different ways: is it actually legal to run an AI-generated image in a paid ad, on a landing page, or in a social feed? The regulators have not written new rules for synthetic imagery, which trips people up, but the existing framework already covers it. If you are commissioning AI creative for a campaign, the team at Absolutely AI works to the same compliance guardrails a traditional shoot would.
The short answer
AI-generated images are legal in Australian advertising in 2026, full stop. What is not legal is using them in a way that misleads consumers, infringes copyright, breaches privacy, or violates a sector-specific code. The technology is neutral in the eyes of the ACCC; the claims and depictions the image supports are what get scrutinised. That is the same standard Photoshop composites have lived under for two decades.
- Legal: AI imagery used honestly, licensed correctly, and disclosed where a platform or code requires it.
- Risky: AI imagery that stylises a product, uses a synthetic person, or leans on prompts referencing a named artist or brand.
- Illegal: AI imagery that misrepresents product features, fakes testimonials, resembles a real person without consent, or copies a protected work.
The four laws that actually apply
There is no AI Ad Act. What you are complying with is a stack of existing statutes and codes, each with its own regulator and its own test. Miss one and the fact that a machine produced the pixels will not save you. The same standards that govern stock photography apply to synthetic imagery, only the failure modes are different.
Australian Consumer Law, section 18
Section 18 of the ACL prohibits misleading or deceptive conduct in trade. It is technology-agnostic: if an AI image gives a consumer a false impression of what they are buying, the ACCC can act. Penalties reach the greater of $50 million, three times the benefit gained, or 30% of adjusted turnover for a corporation, and $2.5 million for an individual. The regulator does not need to prove intent, only that the average consumer would be misled.
AANA Code of Ethics and AI Industry Guidelines
The Australian Association of National Advertisers publishes the Code of Ethics enforced by Ad Standards, plus specific AI Industry Guidelines released to address synthetic media. The guidelines expect advertisers to ensure AI content is truthful, not misleading about human involvement, and reviewed by a person before publication. They do not have the force of law, but a Community Panel determination is embarrassing and platforms often pull the ad.
Copyright Act 1968 (Cth)
Australian copyright requires a human author. A purely machine-generated image may therefore have no copyright protection at all, which means competitors could legally reuse it. Worse, if the model was trained on copyrighted work and the output is substantially similar, you can still be liable for infringement even if you did not know. The Copyright Act does not care that a black box produced the image.
Privacy Act 1988
If your AI image contains a person's likeness that a reasonable observer could identify, the Privacy Act's Australian Privacy Principles are in play. That covers training on real photos, generating a face that resembles a real individual, and any prompt that references a named person. Consent is the safest path, and for talent-adjacent work a signed release is still the standard our brand photography team uses.
When AI images cross the line into misleading advertising
The ACCC has been explicit that generative tools do not change the analysis: if the image misrepresents the product or the experience, it is misleading conduct. The commission has pursued cases where retouched or composited imagery gave consumers a false impression, and AI raises exactly the same issues at a higher volume.
- Fake product features: generating a version of your product with capabilities, sizes or finishes it does not actually have.
- Unrealistic before-and-afters: especially in beauty, fitness and health, where an AI "after" image implies a result the product cannot deliver.
- AI food photography: a burger that looks nothing like what is served is a section 18 problem whether it is styled with props or generated from a prompt.
- Fabricated testimonials and faces: synthetic "customers" holding your product, or AI-generated review avatars, are treated as false representations under ACL sections 18 and 29.
The safe default is that any AI image supporting a product claim should be verifiable against the real product. Our product photography workflow anchors every hero shot to a real reference to keep the depiction honest.
Copyright: who owns an AI image in Australia?
This is the question that most surprises marketing directors. Under the Copyright Act, protection subsists in original works of human authorship. A prompt typed into Midjourney, without meaningful human creative contribution to the output, likely produces something with no Australian copyright at all. That means the exclusivity you assume you are buying may not exist.
The training-data question is the other half. If an AI model was trained on copyrighted images and the output is substantially similar to a specific protected work, using it commercially can infringe. This is why platform terms of service matter as much as the law. Midjourney grants paying users a commercial licence to their outputs, OpenAI grants image ownership to the user under its terms, and Adobe Firefly is trained on licensed and public-domain material with an IP indemnity attached for enterprise customers. Read the tier you are on before you brief a campaign.
Using AI-generated people
A synthetic person is safer than a lookalike, but not automatically safe. If the AI output resembles a real, identifiable individual, you have a passing-off risk, a Privacy Act risk, and in some contexts a defamation risk. Deepfake-style content that puts a real person's face onto a body they did not consent to is squarely in the danger zone, and Meta, TikTok and YouTube already remove it on sight.
The working rule inside a compliant social ad creative pipeline is that any human depicted in a paid ad should be either a real person under a signed release or a synthetic person that has been visually checked against reverse-image search to confirm it does not resemble a real individual. Add a note to file explaining how the image was generated, and keep the prompt.
Do you have to disclose AI use in ads?
As of 2026 there is no Australian law that requires advertisers to label an image as AI-generated. The AANA AI Industry Guidelines lean towards disclosure where the use of AI is material to how a consumer interprets the ad, particularly for testimonials, endorsements and news-style content. That is guidance, not statute.
Platform rules are stricter and they bind you before the law does. Meta requires advertisers to disclose AI-generated or digitally altered content in ads about social issues, elections and politics, and applies a broader synthetic-media label across Facebook and Instagram. TikTok requires creators to toggle an AI-generated content label for realistic synthetic imagery. Google and YouTube require disclosure of altered or synthetic content in election ads and in any ad depicting a real person doing something they did not do. Check the platform's help centre before you push publish, because enforcement is automated.
A 7-point pre-flight checklist before you publish
This is the checklist we run internally on every AI-generated asset before it goes to a client for approval. It takes about ten minutes per hero image and catches almost every issue that would otherwise turn into a complaint or a takedown.
- Claim check: does the image visually support any product claim that is not literally true of the product as sold?
- Likeness check: reverse-image search every AI-generated face; confirm it does not resemble a real identifiable person.
- Brand check: scan the frame for third-party trade marks, logos or protectable trade dress the prompt may have pulled in.
- Licence check: confirm the generation tool's current commercial terms cover paid advertising on the tier you are subscribed to.
- Sector check: apply the relevant industry code (TGA, ASIC, alcohol, food, real estate) before the general codes.
- Platform check: apply the labelling and disclosure rules of every platform the asset will run on, not just the primary one.
- Record check: save the prompt, the tool, the version, the date, and the human sign-off in a compliance folder.
Sector watch-outs
General ad law is the floor. Sector regulators apply their own overlay, and each one has views on synthetic imagery that go beyond the ACL. If you are running AI creative on Meta or any performance channel in one of these categories, budget extra review time.
| Sector | Regulator | What it means for AI images |
|---|---|---|
| Health and therapeutic goods | TGA | Therapeutic Goods Advertising Code bans misleading before-and-afters and unverifiable testimonials, AI or otherwise. Synthetic patient imagery is high-risk. |
| Financial services | ASIC (RG 234) | ASIC's AI-washing focus extends to imagery that implies AI capability or investment performance the product does not have. |
| Food and beverage | ACCC + AANA Food and Beverages Code | Depictions must match the product actually sold; synthetic serving suggestions carry section 18 risk. |
| Alcohol | ABAC | AI-generated drinkers must clearly be over 25 in appearance; no imagery appealing to minors. |
| Real estate | State fair-trading agencies | AI-enhanced property imagery must not misrepresent size, condition or view; some states require disclosure of digital enhancement. |
Frequently Asked Questions
Can I sell AI art commercially in Australia?
Yes, subject to the platform's commercial licence and general law. What you cannot do is assume you own copyright in the output, so treat it as a commissioned asset with a licence, not an asset you can enforce against copiers.
Do I need a model release for an AI-generated person?
Not for a purely synthetic person that does not resemble a real individual. If there is any chance the face resembles a real person, either regenerate or obtain a release. Document the check either way.
Is Midjourney output copyrighted in Australia?
Probably not, because Australian copyright requires human authorship and a prompt alone is unlikely to meet the threshold. Midjourney's terms grant you a commercial licence to use the image, which is the more useful right in practice.
What if the AI copies a real brand or artwork?
You are liable for the infringement even if you did not intend it. Run a reverse-image search on any output that feels stylistically specific, and avoid prompts that name living artists, brands or trade marks.
Do I have to tell customers an image is AI-generated?
Not by Australian law in 2026, but yes on some platforms and yes under the AANA AI Guidelines where the AI element is material. When in doubt, label it; consumers respond better to disclosure than to being caught out.
Can I use AI images in a TVC?
Yes, and CAD clearance still applies. The clearance body will look at the same claims, depictions and sector codes it always has, regardless of how the frames were produced.
The takeaway
Using AI-generated images in Australian advertising is legal, workable and increasingly standard, provided you treat the compliance layer as seriously as you treat the creative. The regulators have not written new rules because they do not need to; the ACL, the AANA Code, the Copyright Act and the Privacy Act already reach every image you publish. Absolutely AI builds the pre-flight checklist into every brief we run through our AI social content workflow, so the creative ships fast and the legal risk stays on the right side of every line.