The ACCC applies the Australian Consumer Law to the outcome, not the tool. AI-generated copy, imagery and testimonials sit under exactly the same s18 and s29 rules as human-written work. Advertisers — not the model — carry the liability. The practical shift in 2026 is that AANA, ACCC guidance and sector regulators now expect you to substantiate, review, and where necessary disclose AI-assisted creative before it goes live.
Marketing teams across Sydney, Melbourne and Brisbane are drafting campaigns with Claude and ChatGPT at a rate that makes any "do we use AI?" policy question obsolete. The question now is how to stay on the right side of the Australian Consumer Law (ACL), the AANA Code of Ethics, and the sector codes that sit on top of them. This guide sets that out.
Australian Consumer Law foundations: s18 and s29
The ACL is Schedule 2 of the Competition and Consumer Act 2010. Two sections do most of the work in advertising matters, and both apply regardless of whether a paragraph was typed by an account manager or generated by a model.
Section 18 — misleading or deceptive conduct. A broad prohibition on conduct in trade or commerce that misleads or is likely to mislead. Intention is irrelevant. If an AI-drafted landing page overstates a feature, implies a partnership that doesn't exist, or uses a phrase that creates a false overall impression, the advertiser has engaged in misleading conduct whether or not anyone read the draft carefully before it shipped.
Section 29 — false or misleading representations. More specific, with civil penalties. It covers false claims about standard, quality, sponsorship, approval, place of origin, price, testimonials and performance characteristics. This is where fabricated reviews, invented awards, and unsupported "Australian made" claims land — all of which large language models will cheerfully generate if you ask them to brainstorm "credibility boosters" without constraints.
ACCC enforcement focus on AI
The ACCC named AI and emerging technologies in its compliance and enforcement priorities across 2024 and 2025, with a specific focus on misleading claims generated or amplified by algorithms. Published statements from Chair Gina Cass-Gottlieb framed the issue plainly: the technology is new, the law is not. Priority areas have included:
- Greenwashing, including AI-generated environmental claims that cannot be substantiated
- Fake or AI-assisted reviews and testimonials, flagged in the Digital Platform Services Inquiry
- Price and discount representations — particularly automated personalisation that implies savings that don't exist
- Influencer content, following the 2023 influencer sweep that found 81% of reviewed posts raised concerns
Testimonials and influencer disclosures
Three rules carry across to AI-assisted work. Testimonials must be genuine and reflect a real customer's actual experience. Commercial relationships must be clearly disclosed — #ad, #sponsored, or plain-English equivalents, not buried in a bio. And an AI-generated or AI-edited quote presented as a real customer's words is a false representation under s29, even if the underlying person is a real customer.
For influencer briefs, ask AI to draft the structure of the post — the disclosure, the call to action, the hashtags — but keep the experiential claims in the creator's own words. If AI is used to generate the spokesperson (a synthetic ambassador), the ACCC and AANA position is that viewers must be able to tell.
AI-generated imagery and video
Synthetic media is where 2026 enforcement is sharpening. The ACCC's concern is not the technology but the impression it creates. A photoreal image of a product that doesn't match the actual product — different colour, different size, fabricated accessories — is misleading conduct. An AI-generated "before and after" that the customer never experienced is a false testimonial.
Current expectations, drawn from AANA guidance and ACCC commentary:
- Label clearly synthetic imagery where a reasonable consumer could mistake it for real
- Do not generate faces of real people, including celebrities or executives, without consent
- Avoid synthetic voices that impersonate identifiable individuals
- Treat AI-generated "demos" of a product as creative concepts, not evidence of performance
AANA Code of Ethics: industry self-regulation
The Australian Association of National Advertisers (AANA) Code of Ethics is enforced by Ad Standards through its complaints process. Section 2 requires advertising to be truthful and honest, not mislead or deceive, and to be clearly distinguishable as advertising. The Code applies to the advertiser, not the production method — so an AI-drafted post that breaches Section 2 is treated the same as a human-drafted one. The AANA's practice notes have specifically addressed AI, confirming that synthetic creative is in scope.
Sector overlays: TGA, ACMA, ASIC
Three sector regulators add obligations on top of the ACL:
- TGA — the Therapeutic Goods Advertising Code governs any claim about medicines, supplements, or medical devices. Testimonials for most therapeutic goods are prohibited. AI-generated health claims need the same evidence file as any other.
- ACMA — broadcast and online content standards, including gambling advertising restrictions and the Online Safety Act framework for synthetic or deceptive content.
- ASIC — the ASIC Act consumer protection provisions mirror ACL s18 and s29 for financial products and services. ASIC's INFO 240 guidance on advertising financial products applies to AI-drafted copy without carve-outs.
State fair trading overlays
State fair trading Acts — NSW Fair Trading Act 1987, Victoria Australian Consumer Law and Fair Trading Act 2012, Queensland Fair Trading Act 1989 — apply the ACL at state level and give state regulators (NSW Fair Trading, Consumer Affairs Victoria, Queensland OFT) their own enforcement powers. For agencies with multi-state campaigns, the practical upshot is that a single misleading ACL claim can attract attention from more than one regulator.
A 5-step compliance framework for 2026
- Set a prompt policy. Decide what AI can and can't be used for: ideation and structure yes; fabricating testimonials, inventing stats, or generating synthetic people no. Store the policy in a Claude Project or ChatGPT custom instruction so every team member starts from the same guardrails.
- Substantiate every factual claim. Before any AI-drafted line about price, performance, savings, outcomes, or environmental benefit goes live, link it to an evidence file. If you can't substantiate it, cut it.
- Human review by someone with sector knowledge. A junior reviewer catching typos is not enough. Finance copy needs an AFSL-aware reviewer; health copy needs TGA literacy; broadcast needs ACMA awareness.
- Label synthetic media. Where an image, voice or video could reasonably be mistaken for real, disclose it. "AI-generated image" or "dramatisation" near the asset is usually enough.
- Keep an audit trail. Log the prompt, the model, the output, the reviewer, and the substantiation source. If the ACCC asks, you want to produce the file inside a day, not a week.