Australia's Teen Social Media Ban Report Used ChatGPT for Editing, Authors Admit

Key points

• A $3.48 million age-verification trial, funded to test technology for Australia's under-16 social media ban, contains citations to academic articles that cannot be found.
• The authors confirmed they used ChatGPT, an AI chatbot, during editing.
• The UK-based Age Check Certification Scheme ran the trial and authored the report.
• Authors deny the missing references are AI hallucinations, the term for when an AI confidently invents sources that do not exist.
• The errors were raised at the Australian Senate.

Australia is pushing ahead with one of the world's strictest social media laws, banning children under 16 from platforms such as Instagram and TikTok. Central to that effort is proving that age-checking technology actually works. So a $3.48 million trial was commissioned to do exactly that.

The problem: the report from that trial appears to cite academic articles that don't exist.

What did the Guardian find?

Guardian analysis found that one section of the report includes references to academic papers that cannot be located. Broken or invented citations matter because policymakers and senators use this research to make real decisions about laws affecting millions of Australian teenagers.

Produced by the Age Check Certification Scheme (ACCS), a UK-based organisation that certifies age-verification products, the report covered a trial run last year testing several technologies social media companies could use to verify a user's age. AI2Day has been tracking age-verification policy since 16 July 2026, and this is the most serious evidentiary challenge to that work we've reported.

Did AI write the dodgy references?

The ACCS authors say no, but they've confirmed ChatGPT was used during editing. That admission matters because AI hallucination, where a chatbot invents plausible-sounding but completely fictional sources, is a well-documented problem with tools like ChatGPT.

Picture asking a very confident assistant to tidy up your bibliography, and that assistant quietly filling gaps with references it made up on the spot. They look real. They're not.

The authors haven't yet provided a clear explanation for why the cited articles can't be found.

What does this mean for the ban itself?

Australia's social media ban is already law. The age-assurance trial was meant to help work out how to enforce it, testing which technologies are reliable enough to check a child's age at sign-up.

Research underpinning those technology recommendations that contains invented sources puts the quality of those recommendations genuinely in question. Senators raised the issue during hearings, so this won't quietly go away.

For parents and teenagers, nothing changes immediately. But the credibility of the evidence base for how the ban gets enforced is now under scrutiny.

What should you take from this?

This is a useful reminder that AI editing tools can introduce errors that look polished and authoritative. If you're using ChatGPT to help write or edit anything containing references, check every single citation yourself before it goes out. AI can't reliably be trusted to keep your sources accurate.

Common questions

What is an AI hallucination?

An AI hallucination is when a chatbot like ChatGPT generates information that sounds credible but is simply made up, including fake quotes or references to papers that don't exist.

Is ChatGPT free to use?

ChatGPT has a free tier giving access to its standard model, and a paid plan called ChatGPT Plus that unlocks faster, more capable versions.

Should you worry about AI errors in government reports?

Reports that inform laws or public policy should always have citations checked by a human expert before publication, regardless of what tools were used to draft them. This case shows why.