A government-backed report that helped pave the way for Australia’s under-16 social media restrictions is facing fresh questions after an analysis found fabricated and inaccurate academic references, some of which were linked to the use of ChatGPT.
The report, which cost A$3.48 million, was produced as part of a government-funded trial examining technology that could be used to verify the ages of people accessing online platforms.
However, its credibility has now come under pressure after researchers examining the report identified multiple citations that appeared to be incorrect, non-existent or unrelated to the claims they were being used to support.
The findings raise broader concerns about the use of generative AI in government research, particularly when such research is used to inform major public policy decisions.
Report’s AI use becomes a point of contention
The Age Check Certification Scheme (ACCS), which conducted the age-assurance technology trial, initially said that artificial intelligence had not been used to produce the report.
That position later changed.
ACCS acknowledged that ChatGPT had been used to rewrite sections of the document after metadata associated with the AI tool was discovered in four links across two sections of the report.
The organisation maintained that the presence of the metadata effectively disclosed the AI involvement and said the references in the report had been manually checked.
But an analysis by Guardian Australia identified six problematic references in the report’s chapter on emerging technologies.
The issues went beyond simple formatting mistakes.
Researchers find non-existent papers and misleading citations
Several Digital Object Identifiers, commonly known as DOIs, were found to point to academic papers that did not appear to exist.
In other cases, the information associated with the DOI included combinations of authors, journals and publication dates that could not be verified elsewhere in the academic literature.
One citation did lead to a genuine research paper, but the paper did not support the claim attributed to it in the government report.
Such errors are particularly significant in a report intended to provide evidence for policies affecting millions of young Australians.
The problems also appeared to continue after the initial concerns were raised.
Attempts to correct some of the disputed citations reportedly resulted in additional inaccuracies. In one case, an ACCS source claimed that a paper had been accessed in March 2025, even though the paper was not actually published until June 2025, according to its lead author.
Concerns over policy decisions
The controversy has prompted questions about how much confidence policymakers should place in research containing unreliable references.
Australian National University academic Christian Downie warned that false references can undermine both policymaking and public trust, regardless of whether they were generated by AI or introduced through another process.
The issue is particularly sensitive because the report played a role in the policy process surrounding Australia’s controversial decision to restrict social media access for children under 16.
Communications Minister Anika Wells had previously praised the report for identifying potential solutions around age verification.
The minister’s department also told a Senate inquiry that it had met with ACCS and that the organisation had explained the problems involving faulty links.
However, the department was reportedly not informed that some of the citations had been fabricated.
AI accuracy becomes a bigger government concern
The controversy is not occurring in isolation.
Independent Senator Fatima Payman compared the situation with a separate case involving Deloitte last year, when the consulting firm refunded part of a A$440,000 government contract after a report was found to contain faulty references generated using AI.
The latest case has therefore reignited a much broader question: how should governments and research organisations use generative AI when producing evidence that could influence public policy?
AI tools such as ChatGPT can help researchers draft, summarise and organise information, but they can also generate convincing-looking citations that do not exist or do not support the claims attached to them.
That makes independent verification particularly important when AI-assisted material is used in official reports.
For Australia, the controversy could now add another layer of scrutiny to an already contentious under-16 social media policy — with questions not only about the ban itself, but also about the reliability of some of the evidence used to help justify it.
