Do You Have to Label AI-Generated Marketing Content, Images, or Ads as AI-Generated?

Detailed Explanation

"Do we have to say this was made with AI?" comes up constantly as businesses adopt AI for marketing content, but the honest answer splits into two very different obligations depending on what kind of content is involved.

Most AI-assisted marketing content carries no AI-specific labeling requirement in Australia. A product description drafted with Claude or ChatGPT, a social media caption, a blog post outline turned into a full article — none of these currently trigger a general "this was AI-generated" disclosure law, as long as the actual claims made are truthful. The legal question that's always applied to marketing copy under the Australian Consumer Law — is it accurate, is it misleading — still applies exactly the same way whether a person or an AI tool drafted the words. Using AI as a writing tool doesn't create a new disclosure obligation on its own.

One category carries real, existing obligations regardless of any AI-specific law, and it's where the actual compliance risk sits for an Australian business: fake endorsements and testimonials. Whether or not any AI-specific law applies, the ACL's misleading-or-deceptive-conduct provisions — enforced by the ACCC — already prohibit presenting a fabricated review, testimonial, or endorsement as if it came from a genuine customer. An AI-generated "customer" saying they love your product is a deceptive-endorsement problem first, and the ACCC has flagged fabricated AI content and "AI-washing" (misleading claims about AI involvement or capability) as an active enforcement focus.

A separate category applies only if your business has EU customers or users: synthetic image, audio, and video content that could pass as real. The EU AI Act's transparency provisions include a labeling requirement for AI-generated or manipulated content — image, audio, or video — that appreciably resembles real people, objects, places, or events and could reasonably be mistaken for authentic. This is commonly discussed as the Act's "deepfake" rule, and it's a distinct obligation from the chatbot-disclosure rule covered in do you have to tell customers they're talking to an AI chatbot — that page covers live conversational interactions; this covers published or pre-recorded synthetic media. It doesn't apply to a purely domestic Australian business with no EU-facing content or audience.

Australian Consumer Law and Fake Reviews

Regardless of any AI-specific statute, the ACL's prohibition on misleading or deceptive conduct (enforced by the ACCC) already covers advertising practices that misrepresent whether a review, testimonial, or endorsement reflects a genuine customer's honest experience. Generating a synthetic "five-star review" or a fabricated customer testimonial with AI doesn't create a new category of risk so much as it applies an existing rule to a new production method — the ACCC treats fake or AI-fabricated reviews as squarely within its existing enforcement authority, and penalties under the ACL have increased significantly in recent years, giving this real teeth.

The practical distinction that matters here: AI-assisted drafting of your own business's honest marketing claims is a different situation from using AI to fabricate the appearance of a third party's endorsement. The first is a writing tool; the second is the deceptive practice the ACCC is actually focused on.

The EU AI Act's Content-Labeling Rule (Only Relevant With EU Exposure)

If your business has EU customers, users, or audiences, Article 50 of the EU AI Act covers more than the chatbot-disclosure obligation — it also requires deployers to mark AI-generated or manipulated audio, image, or video content that could pass as authentic, and, in some circumstances, AI-generated text published to inform the public on matters of public interest. Both obligations become applicable from 2 August 2026, alongside the chatbot rule. A business with no EU-facing marketing can treat this section as background rather than a compliance requirement.

Two details matter in practice for a business this does apply to:

• The threshold is "could pass as authentic," not "was made with AI." A visibly stylised or clearly synthetic marketing graphic is a different case from a photorealistic synthetic video presented without any indication it's generated — the rule targets content designed or likely to be mistaken for something real, not AI-assisted creative work generally.
• This applies to your business as the deployer of the content, not only to whichever AI tool generated it. The same "deployer responsibility" pattern that applies to the Act's chatbot-disclosure rule applies here — see does the EU AI Act apply to a business using ChatGPT or Claude for that distinction in full, and how it compares to the Australian Consumer Law obligations that apply to every Australian business regardless of EU exposure. As with the rest of the Act's newer provisions, confirm the current implementation guidance and any phased timelines before relying on a specific detail here.

What This Means for Ordinary Business Marketing

For most small and mid-size businesses using AI to help write and design marketing content, the practical takeaway is narrower than the headlines about AI-content regulation suggest:

• Routine AI-assisted copywriting — product pages, ad copy, social posts, email marketing — needs no AI-generated label under Australian law, as long as the claims made are accurate. Keep applying ordinary honest-advertising standards under the ACL, not a new AI-specific checklist.
• Photorealistic synthetic images, video, or audio presented as real — a synthetic "customer" testimonial video, a fabricated "behind the scenes" clip, an AI voice presented as a real person's recorded endorsement — is where the ACL's deception rules apply regardless of your business's footprint, and where the EU AI Act's labeling rule adds a further, separate obligation if you have EU exposure. A clear "AI-generated" or "digitally created" label is the safer default either way.
• AI-generated reviews or endorsements attributed to real or implied real customers are a bright-line no under the ACL, independent of any AI-specific statute — this was already prohibited before AI made it easier to produce at scale, and it's an active ACCC enforcement focus.

Things to Consider

• This is a different obligation from chatbot disclosure — check both, not just one. A business running both an AI chatbot and AI-generated marketing visuals has two separate transparency obligations to satisfy, covered by two different parts of the same regulation; see do you have to tell customers they're talking to an AI chatbot for the live-interaction side.
• "Could pass as authentic" is a judgment call for genuinely realistic content, not routine marketing graphics. Most AI-assisted marketing content — obviously stylised images, clearly AI-drafted copy — doesn't approach the threshold this rule targets; the risk concentrates in photorealistic synthetic media presented without any indication of its origin.
• Honest advertising law doesn't go away because AI was involved. Whatever a human-drafted ad would need to substantiate, an AI-drafted version needs to substantiate too — AI doesn't create a compliance shortcut or a new loophole in either direction.
• This is a genuinely fast-moving area. ACCC guidance on AI-generated content and "AI-washing," the EU AI Act's content-labeling implementation guidance, and a growing number of jurisdiction-specific synthetic-media and deepfake laws (some aimed narrowly at elections or explicit content, some broader) are all still developing as of mid-2026 — treat any specific detail here as needing a re-check against current sources, and get a lawyer's opinion before a specific high-stakes campaign relies on an interpretation of a borderline case.

Common Mistakes

• Assuming any use of AI in marketing now requires a disclosure label. This overcorrects into unnecessary "AI-generated" labels on ordinary copy that current rules don't actually require, while sometimes missing the narrower cases — synthetic realistic media, fake endorsements — that carry real obligations.
• Treating a synthetic testimonial as a copywriting shortcut rather than a deceptive-endorsement risk. A fabricated customer quote or review is a compliance problem regardless of how it was produced — AI just makes it faster to generate at scale, not safer to use.
• Conflating this with the chatbot-disclosure rule. Telling a customer they're chatting with an AI, and labeling a piece of synthetic published media, are two different Article 50 obligations with different triggers — satisfying one doesn't satisfy the other.
• Not revisiting the question as guidance develops. This is one of the least settled areas of AI-content regulation as of mid-2026 — a labeling approach that felt sufficient earlier this year is worth checking against current implementation guidance, not assumed to still be correct.