I studied how rumours and misleading information spread on X during the Voice referendum. The results paint a worrying picture

When the Australian public voted last year on whether to change the Constitution to establish an Indigenous Voice to parliament, it came after months of intense and sometimes bitter campaigning by both the “yes” and “no” camps.

Polling conducted 12 months before the referendum showed majority public support for the proposed constitutional change. But ultimately the polls flipped and 60.06% of Australians voted “no”.

Why? Factors included a lack of bipartisan support, a growing distrust in government, confusion about the proposal’s details and enduring racism in Australian society.

However, my research , published in the journal Media International Australia, highlights how misinformation and conspiratorial narratives on social media platforms – in particular, X (formerly known as Twitter) – also played a key role.

The findings paint a striking picture. There is a new type of political messaging strategy in town – and it needs urgent attention.

A bird’s-eye view of campaign messaging

I collected 224,996 original posts on X (excluding reposts) containing search terms relevant to the referendum (for example, “Voice to Parliament” or #voicereferendum). The data collection spanned all of 2023 up to the referendum date on October 14. It included more than 40,000 unique user accounts.

First, I categorised posts based on the presence of partisan hashtags. This enabled the identification of the top 20 keywords associated with each campaign.

The results provide an aerial view of each campaign’s messaging strategy. They also reveal that keywords associated with the “no” campaign dominated on the platform.

Keywords from the “yes” campaign included, for example, “constitutional recognition”, “inclusive”, “closing the gap” and “historic moment”.

Keywords from the no campaign included, for example, “division”, “expensive”, “bureaucratic”, “Marxist”, “globalist” and “Trojan”.

I found the “no” campaign keywords occurred more than four times as often in the dataset as the “yes” campaign’s, with the “not enough details” and “voice of division” narratives most prevalent of all.

The “yes” campaign only had two of the top ten keywords in campaign messaging on the platform.

How did the ‘no’ campaign manage attention on X?

I categorised each post in the dataset according to its dominant theme or topic. The top ten most prevalent topics covered the majority of the dataset (64.1% of all posts).

Next, I examined which of the top ten topics gained most attention on X – and which X users were the most influential.

Across the board, the posts that received the most engagement (that is, the number of replies and reposts with an attached original message) were from politicians, news media and opinion leaders – not bots, and not trolls.

In line with the keyword analysis, the “no” campaign messaging dominated the topics of discussion, but not because everyone agreed with it.

Several of the topics featured core “yes” campaign messaging, emphasising First Nations representation and equality, opportunities to make a difference and historical facts.

But most of the discussion from “yes” campaigners was drawing attention to and critiquing the “no” campaign’s core messaging around fear, distrust and division.

Rather than blatant falsehoods or full-blown conspiracy theories, the most widely discussed posts from “no” campaigners were characterised by rumours, unverified information and conspiratorial assertions.

Prominent “no” campaigners portrayed the Voice as divisive, implying or arguing it would lead to drastic social changes such as apartheid. It was positioned as part of an alleged secret agenda to consolidate elite privilege and erode Australian democracy through risky constitutional changes.

Such claims are indisputably conspiratorial because they assert that powerful actors are hiding malevolent agendas, and because they lack credible and verified empirical evidence.

These claims were supported by collaborative work by No campaigners to find what they believed to be evidence.

This type of “just asking questions” and “ do your own research ” approach stood in contrast to the journalistic fact-checking and traditional expertise predominantly drawn on by the “yes” campaign.

Yet, my study’s results show that the more the “yes” campaign tried to counter misrepresentations and confusion around the Voice proposal, the more they fuelled it.

A post-truth referendum

What Australia witnessed in October 2023 was a thoroughly post-truth referendum.

To be clear: it was not a referendum that lacked truth, but one in which traditional political messaging simply didn’t cut it in a fast, free-flowing and predominantly online media environment.

The “no” campaign’s messaging strategy was all about constructing a “truth market” in the public sphere. In other words, they created an environment where multiple – often conflicting – versions of the truth competed for dominance and where emotional resonance received more attention than reasoned debate.

We can’t really call it “disinformation” because most of it didn’t involve outright falsehoods.

Instead, a near-constant supply of contrived media events and rumour bombs attracted 24/7 news attention and fostered participatory discussions on platforms such as X from actors across the partisan divide. Examples of rumours spread during the campaign included the Indigenous Voice to parliament would divide Australia and was a land grab for globalist elites.

In trying to counter the “no” campaign’s messaging on X, many “yes” campaigners entered into a “ defensive battle ”. This drowned out their core message. It also amplified the fear and division narratives of the “no” campaign.

It’s a classic example of the oxygen of amplification .

Targeted messaging designed to exploit social media and elicit reliable outrage from different segments of the population is not new. It has a name: propaganda .

Propaganda is not a bad word, despite the reputation it has developed since the second world war. It is simply a more accurate and principled way to understand what happened during the Voice referendum debate, and for political campaigning more generally.

What is new, however, is the current information environment: the speed of digital networks and the collaborative and social dimensions of how people engage with information.

Properly diagnosing the problem is the first step to remedying it.

Timothy Graham receives funding from the Australian Research Council (ARC) for his Discovery Early Career Researcher Award (DE220101435), 'Combatting Coordinated Inauthentic Behaviour on Social Media'. He also receives ARC funding for the Discovery Project, 'Understanding and combatting "Dark Political Communication".' (2024 - 2027).