August 31, 2026, (Inside AI) — Dozens of policy submissions to Australia’s federal parliament contain fabricated academic references and invented research findings, according to a new analysis. The documents, spanning multiple inquiries and political perspectives, incorrectly attribute studies to real scholars and cite papers that do not exist.
The findings point to a growing problem: large language models are flooding democratic processes with plausible-looking but false evidence. In some committee reports, the majority of cited sources appear to be AI hallucinations.
Christian Downie, a professor at the Australian National University’s school of regulation and global governance, warns the stakes are high. If the trend continues, parliamentarians risk “making decisions based on evidence that doesn’t exist.”
Fake Citations Undermine Expert Trust
One submission to an inquiry on family violence and suicide included a hallucinated reference attributed to Divna Haslam, an associate professor and clinical psychologist at the University of Queensland. The fake citation misstated her team’s research findings.
Haslam called the reference “scary” because it looked so real. Google’s AI summary even summarized the fake paper as if it were genuine.
“It’s very frustrating to ... to have invested time, money, effort, expertise in rigorous research,” she said. “Then to see something that’s inaccurate and inappropriately attributed anyway is really concerning.”
The submission’s author, Drilldown Reports, acknowledged using AI in its research process. A spokesperson said the company identified the errors in a follow-up submission but missed the correction deadline.
“We do strongly believe in the human factor and that it should be a vital part of the full quality review process,” the spokesperson said. “Unfortunately, the human factor also lets us down when uploading the correct document.”
The spokesperson rejected the framing that AI was to blame: “I get that you feel you have a juicy article to add the AI fear machine but this was a human error, not AI.”
Dozens of Submissions Contain Fabricated Sources
The analysis identified at least 39 submissions containing what appear to be hallucinated references. The figure is a conservative estimate because it only catches documents with incorrect citations, not AI-generated text without references.
More than 100 papers included ChatGPT URL tags in reference links, a sign that the tool was used during drafting. The flawed submissions came from individuals and organizations across the political spectrum.
When contacted, many authors said they were unaware AI could invent references or admitted they had missed the errors before submitting.
The problem extends beyond parliament. Last year, consultancy giant Deloitte issued a partial refund to the federal government after AI tools produced fake references, including a fake court citation, in a $440,000 report.
Downie says the implications go beyond bad decisions. “We’re also starting to undermine the public’s trust and confidence in the types of institutions that underpin our democracy,” he said.
Nicole Gurran, a professor of urban and regional planning at the University of Sydney, found fake citations attributed to her in a housing inequity inquiry submission. She said fake citations break the chain of evidence, even when the underlying claim is correct.
“Fake citations, even if an accidental ‘collage’ where the claim is correct but the chain of reference to evidence is broken, undermines that,” she said.
Margaret Simons, a journalist and academic, discovered a submission to a Senate inquiry attributed a claim about the Guardian to an article she never wrote. The endnote was so convincing that Simons briefly questioned herself.
“Even though I know I didn’t write this, I did have that moment of self questioning,” she said.
OpenAI said addressing hallucinations is an ongoing research area. Users should treat ChatGPT as “a first draft, not a final source” and verify all quotes, data, and references.
Google said AI Overviews operate like traditional search, matching web content to query terms. The company said the feature surfaces pages that include those terms, just like standard blue link results.
Downie says new guidelines may be needed to “encourage truthfulness” in the inquiry process. “Whether it’s climate change, immigration, health, we want our elected officials to be making decisions based on real information and real evidence not on fake citations and fake claims,” he said.
