84% of Australians want AI content labelled by law
The short answer
Deloitte's 2026 Media & Entertainment Consumer Insights report, released on 3 November, surveyed Australians about AI content and trust. Three in four said AI-generated content makes it harder to know what is real online. 84% said platforms and publishers should be legally required to label it. 71% said they would stop using a platform that failed to disclose it clearly.
What the survey found
The report is Deloitte's fifteenth annual edition of the study, and this year the AI questions moved from the edges to the centre. Alongside the labelling numbers, the survey found that 57% of respondents say AI-generated search summaries make searching easier, and that those summaries prompt them to search more, not less.
That combination is the interesting part. The same people who find AI summaries useful are the ones who say they can no longer tell what is real. Convenience and distrust are arriving together, in the same sessions, on the same screen.
Deloitte also reports that Australians trust AI content more when they believe humans were involved in making it. Trust, in other words, is tracking visible human responsibility rather than the technology itself.
Why labelling is the ask, not banning
It is worth noticing what the 84% are actually asking for. Not fewer AI summaries, not a ban on generated images. A label. That is a modest ask, and it is consistent with how people handled earlier waves of synthetic media: they want to know what they are looking at, then decide for themselves.
The 71% figure is the harder one. A majority says it would abandon a platform that did not disclose AI content clearly. If that number is real purchasing behaviour rather than a survey answer, disclosure stops being a compliance cost and becomes a retention feature.
Where this connects to policy
Australia has been moving on AI transparency for a while, and this survey gives regulators a public-opinion number to point at. Mandatory labelling rules are already part of the debate in the EU under its AI Act transparency provisions, and platforms have been adding AI labels and content credentials on their own, unevenly, in the meantime.
The gap the report exposes is between voluntary labels and what people say they need. A label that appears on some images, in some apps, sometimes, is not the same as a rule. That gap is exactly where the argument will be fought.
What it means if you publish anything
- Label at the point of publication, not in a settings menu nobody opens. The survey's complaint is about not knowing, and that is a placement problem as much as a policy one.
- Keep a human in the loop on anything factual. Respondents trusted content more when they believed a person was involved, so say who checked it.
- Expect the rule to arrive. If mandatory labelling lands, publishers with a labelling habit already in place will spend a week on compliance. Everyone else will spend a quarter.
The number to remember
Three in four. That is how many Australians say AI content has made it harder to tell what is real. It is not a fringe complaint from people who avoid the technology. It is the majority view of a general population sample, released the same week that AI summaries became the default way a lot of people read search results.