Final Data resources-1

What the Marketing Association expects from 2027

Learn what the ASA's updated AI advertising guidance means for New Zealand marketers, including disclosure requirements, misleading content and advertiser responsibilities from 2027.

Last month the Marketing Association published a completely new Code of Practice for Marketing Communications which included an entire section on best practice use of AI. At the same time the ASA published an updated Advertising Standards Code and, beside it, a Guidance Note on AI use in advertising. The Code applies to new advertising from 1 February 2027 and to all advertising within the ASA's jurisdiction from 1 April 2027.

Both codes are principles-based and remarkably similar. The existing rules on truthful presentation and substantiation apply when AI is involved. Marketers still have to answer the same question as before, which is whether the ad could mislead someone. AI makes that easier to do by accident, and at volume.

The most common thing I come across is marketers who have used AI to create part of a social post or a graphic for their website and are unsure what applies to them on disclosure. I encourage them to err on the side of transparency, and to make sure whatever was generated is actually good quality and checked.

You don't have to label everything

Our industry organisations don’t expect every message made with AI tools to carry a label. The test is whether the AI changes what a consumer understands. Consider a label when AI is used to:

  • create a realistic person who does not exist
  • copy a real person's likeness or voice
  • show events, experiences or results that did not happen
  • produce product imagery that differs materially from the product
  • present anything a consumer would reasonably assume is real

A label is unlikely to be needed for lighting and quality fixes, AI-drafted copy that describes the product accurately, decorative backgrounds, obviously fantastical scenes, or tidying up a genuine testimonial without changing what it says.

Where you do label, keep it plain and put it next to the content it describes. The ASA's examples include "AI-generated image", "AI-generated person" and "AI-generated voice". On-screen text, a caption, audio or accompanying copy all work, as long as the prominence suits the medium.

A label won't rescue a misleading ad

Disclosure is described as "a transparency measure, not a remedy for misleading advertising". A good example is a skincare before-and-after showing results the product cannot achieve. Adding "AI-generated image" to that does not make it compliant. The image has to change.

Testimonials follow the same logic. AI can help you edit, summarise, translate or present a real customer's words, provided the meaning holds and you can substantiate the claims. Keep the original feedback on file. AI cannot invent the customer, the review or the experience.

Every asset lands in one of three places: it needs no label, it needs a label, or it needs fixing because no label will help.

Responsibility stays with the advertiser

You are responsible for your communications whether the AI content came from your team, your agency, a freelancer or a platform's built-in tools. The same applies when AI personalises, targets or distributes the ad. If targeting changes who sees it or how they read it, that is yours too.

The practical consequence is that you need to know where AI was used in a finished asset. That means asking suppliers and recording the answer.

The guidance note gives five questions to ask before publishing:

  1. Could consumers reasonably believe the AI content is real?
  2. Could it affect how they understand the product, service, person or event?
  3. Could the ad mislead through omission, implication, or overall presentation?
  4. Is every objective claim supported by evidence?
  5. Would disclosure help prevent misunderstanding?

A yes to any of them means stopping to check the Code and decide whether a label is appropriate.

Two caveats. The note is guidance. It does not bind the Complaints Board, which judges each ad on the overall impression a consumer is likely to take from it. And the Fair Trading Act applies regardless of how an ad was made. Therapeutic, financial, alcohol and other regulated categories carry their own codes on top.

A two-minute check

AI Innovisory have built a free AI Disclosure Checker to go with the ASA guidance note. It asks up to nine questions about a single ad, post, video or campaign asset. It tells you whether a label is likely to be needed, suggests wording, and says so when a label won't solve the problem. It also covers chatbots and digital twins, which the ASA note does not address, using the IAB's voluntary disclosure framework. It is guidance, not legal advice.

Before February, do two things. Add the five questions to your approval process, with one named person who owns the call. Then look at what is already live and will still be running on 1 April.


Author: Peter Mangin is the founder of AI Innovisory and specialises in AI Enablement and is a member of the AI, Measurement, Privacy and Data Thought Leadership Group, 17th November 2026.