If AI-written marketing claim turns out to be false advertising, who is responsible?
False advertising is a marketing claim that is false or misleading, and under Australian Consumer Law the business that publishes it is liable, not the AI tool that drafted it. Using AI does not transfer responsibility. It only changes who types the first draft, so the claim is still yours to defend.
Picture the moment. You have an AI draft in front of you, it reads well, and one line promises your clients a specific result in a specific number of days. You did not write that number. You are not sure where it came from. Do you publish it? That small hesitation is the whole problem, and most owners push past it because the copy sounds confident.
Does using AI move the liability off the business?
No, and this is the part worth being clear about. The AI is a tool, not a party. Australian Consumer Law looks at who published the claim in trade or commerce, and that is the business, not the software vendor. Treasury’s own 2025 review concluded the existing law already covers AI conduct, so there is no separate AI shield to hide behind.
The stakes changed this year. From 28 March 2026 the maximum penalty for a corporation that breaches the Australian Consumer Law rose from AU$50 million to AU$100 million per contravention. The maximum is the greater of AU$100 million, three times the benefit obtained or 30% of adjusted turnover for the breach period. And section 18 can be contravened without any intention to mislead. An honest mistake is no defence.
| Myth | Reality |
| The AI wrote it, so the AI provider is liable | The business published it in trade or commerce, so the business is liable |
| No intent to mislead means no breach | Section 18 can be breached without intent; an honest mistake is no defence |
| Australia needs a new AI law before this applies | Treasury’s 2025 review found the existing Australian Consumer Law already covers AI conduct |
| Small businesses are too small to be pursued | Penalties scale to the business, and the obligation applies to every advertiser regardless of size |
How do you check an AI-written claim before you publish it?
You run a three-point review on every draft, every time. It takes minutes and it is the whole protection.
- Isolate the factual claims. Read the draft and mark every line that states something a customer could rely on: a number, a result, a guarantee, a comparison, a timeframe, an “Australian made” or similar origin claim. Opinion and description are not the risk. Stated fact is.
- Substantiate each marked claim against a source you hold. For every marked line, point to real evidence you could show the ACCC on request: a record, a result, a supplier document. If you cannot substantiate it now, you cannot publish it. The ACCC can issue a substantiation notice requiring exactly this.
- Check for the claims AI invents. AI tools state confident specifics that were never in your inputs, a percentage, an award, a “number one” line. Delete or verify any claim you did not give the tool, because the business owns it the moment it is published.
Here is the check running. An AI draft for a coaching business says “our clients see a 40% increase in enquiries within 30 days”. Step 1 marks it as a factual, time-bound result claim. Step 2 asks for the records behind 40% and 30 days. The owner has two client cases, not a measured average, so the claim is not substantiated. Step 3 confirms the 40% figure was never in the brief. The AI produced it. The line is cut and replaced with a specific, provable statement about the two named outcomes. The post publishes with nothing the business cannot defend.
Copy the prompt below, paste it in with your draft, and read what it flags.
Read this draft and find every factual claim in it, every number, percentage, result, date, guarantee, ranking, and “made in” style origin claim. For each one, tell me: is this backed by a real, citable source, or did you generate it without one? List any claim you cannot tie to a genuine source, and show me the source where you can. Do not fix the draft, just show me what needs checking before I publish.
Anything on that flagged list is a claim you now check against evidence you actually hold, and cut or rewrite if you cannot back it. That is the three-point check, done in a couple of minutes, on your own work.
In Australia the ACCC enforces the Australian Consumer Law and can take false advertising to the Federal Court, alongside state consumer agencies like Fair Trading. It can also issue a substantiation notice, requiring you to produce the evidence behind a claim. From 28 March 2026 that maximum corporate penalty per breach doubled.
This is not a grey area either. The ACCC has named “AI-washing”, misleading claims about AI capability, as an enforcement priority for 2026 to 2027. AI-related claims are under active scrutiny, so the review check is how you meet an obligation you already hold, not extra caution.
The AI company wrote it, so shouldn’t they be liable?
It feels fair, but the law does not read it that way. Under Australian Consumer Law the conduct in trade or commerce is the business’s publication of the claim, so responsibility follows the publisher. The customer deals with you, not the AI provider. Treasury’s 2025 review deliberately kept it this way, finding the current law adequate. Which means the human review before publishing is the one thing standing between a confident AI line and a claim you cannot back.
Frequently Asked Questions
What counts as false advertising under Australian Consumer Law?
A representation made in trade or commerce that is false, misleading or likely to mislead. Under section 18 it can be breached without any intent to mislead, so an honest mistake is no defence, which is why every claim needs substantiating first.
How do I make sure an AI-written claim is safe to publish?
Run the three-point review on every draft: isolate the factual claims a customer could rely on, substantiate each against evidence you actually hold, then delete or verify anything the AI stated that you never gave it.
What is the difference between an opinion and a claim in marketing copy?
Opinion and description carry no real risk. A stated fact does. A number, a result, a guarantee, a comparison, a timeframe or an origin line is something a customer relies on, so that is the line you must be able to prove.
Can I be liable if the AI invented a statistic I did not check?
Yes. Once published, the business owns the claim. In the worked example an AI produced a 40% result figure that was never in the brief. Because it could not be substantiated, the line had to be cut before publishing.
What results can I expect from a pre-publish review check?
Nothing gets published that you cannot defend with evidence you hold. The check catches the confident specifics AI invents before they reach a customer, which is how you meet an obligation you already carry as the publisher.
How long does the three-point check take?
Minutes per draft. You read it once to mark the factual claims, match each to a record or result you hold, then scan for any specific the tool added on its own. It is a quick habit, not a legal review.
At AI Mastering Hub every piece passes through an AI Marketing Team review and revision layer before it reaches the client, which is the three-point check built into how content is produced. If you want the review check as something you can run yourself, we can walk you through it and show you where AI most often invents a claim. If that is useful, book a 15-minute discovery session and bring a draft you are unsure about.
REFERENCES:
- Doubling of ACCC penalties from 28 March 2026: Treasury Laws Amendment (Doubling Penalties for ACCC Enforcement) Act 2026, reported by ACCC Fines and penalties, https://www.accc.gov.au/business/compliance-and-enforcement/fines-and-penalties and Gilbert + Tobin, 2026
- Section 18 no-fault liability: Australian Contract Law, https://www.australiancontractlaw.info/legislation/acl/s18
- Treasury review of AI and the Australian Consumer Law, Final report, October 2025: https://treasury.gov.au/sites/default/files/2025-10/p2025-702329-fr.pdf
- Business responsibility for AI outputs and ACCC AI-washing enforcement priority: Prosper Law, 2025, https://prosperlaw.com.au/is-your-ai-marketing-exposing-you-to-liability/ ; MinterEllison, 2025
- Substantiation-notice power: ACCC, https://www.accc.gov.au/consumers/advertising-and-promotions/false-or-misleading-claims