This week, researchers published an analysis of more than three million posts and news items about anti-First Nations hate during the 2023 Voice referendum, Anzac Day 2025, and the so-called March for Australia rallies. The numbers are stark. Hateful posts climbed a staircase through the Voice campaign to about 3,600 a day on the eve of the referendum, and by 2025 white supremacist claims and dehumanising language made up almost ninety per cent of what was detected.
I am not going to walk through the findings. Aboriginal and Torres Strait Islander people do not need them walked through. The study is careful, necessary and welcome, and it tells this country what we have always known, because we live inside it. What I want to write about is why the country needs to be told at all.
The racial imagination
What three million posts record is not an internet problem. It is the racial imagination of Australia, the settled repertoire of stories the nation tells about who belongs, who deserves, and whose experience counts, surging into speech at every moment its ownership of the country is touched. Look at when the hate rises. A referendum on recognition. A Welcome to Country – booed again this year, while the political response reached for every word except racism. A day commemorating arrival. Aileen Moreton-Robinson named the twin logics decades ago: the possessive logics of patriarchal white sovereignty, forever reaffirming the nation as a white possession, and the discourse of pathology, which renarrates the harm done to us as evidence of what we are. I have argued across this series that Australia has spent two centuries imagining Aboriginal people as a problem to be managed, a gap to be closed, a population in deficit. This study is that imagination, weighed.
The country has been asked whether it sees us directly – once, at a ballot box. In October 2023, Australians were asked whether the Constitution would recognise the First Peoples of this country and hear them through a Voice. Sixty per cent said no, while the hateful posts crested at three thousand six hundred a day. A No vote was not itself hate – the study was careful about that, and so am I; First Nations people had their own principled critiques. What the annotators coded was the story that carried the campaign: recognition portrayed as racial oppression of non-Indigenous people, our being seen framed as other Australians losing. That story won. The referendum did not create what this data records. It measured it.
The blindness
The study has done the weighing. What follows is the part it cannot do for the country. Most of the country will not recognise the portrait, and the failure is not an information gap. White Australia does not see racism towards Aboriginal people, because it has never fully seen us as people. Terra nullius was never only a claim about land. It was a claim about persons a continent declared empty because the peoples in it did not register as peoples whose law, whose presence, whose injury counted. Under that description there is nothing to see: no racism, only a problem population to be managed. And the blindness is not a metaphor. A decade of implicit association testing across more than eleven thousand Australians found around three in four hold negative implicit attitudes toward us – and most do not know they do.
The proof sits inside the study’s own data. Among the language First Nations community annotators identified as hateful was language denying the historical and ongoing effects of colonisation and racism. Denial is not the absence of the hate. It is one of its genres. The imagination that produces the abuse produces the inability to see it – one faculty, doing its two jobs.
And the blindness has a twin. We have been saying this. First Nations people and organisations have said it at this inquiry’s own hearings. Call It Out logs around five hundred reports a year. Aboriginal women put formal recommendations on violence and abuse to national conferences and taskforces from 1980 onward and were answered with silence for a quarter of a century until a white Crown Prosecutor said the same things on national television, and an army was dispatched within the year – the 2007 Northern Territory Intervention. I set this record out in March, in the first piece of this thread. The testimony was never missing. The ear was never turned.
And the blindness does not maintain itself. What keeps it in working order is the politest mechanism of all: in Australia it is not done to talk about race. Naming racism is received as the disruption, ruder than the thing it names. And for us the cost is specific: every Aboriginal person who has sat in an Australian meeting room knows the calculation: say the word and become the angry Black man, the angry Black woman, the problem in the room. So we pre-edit. We say culturally unsafe when we mean racist, challenging when we mean hostile. The self-censorship is not timidity. It is an accurate reading of the price list. And it is why the online record matters: the comments section is the one place the imagination speaks without its manners on. The daytime discourse of harmony and the three thousand six hundred posts a night are not two Australias. They are one imagination in its two registers – and the polite register has spent decades ensuring that neither is called racism.
I have spent a career inside this definitional question leading the research for Diversity Council Australia’s Racism at Work report and responding to the Australian Human Rights Commission’s Respect at Uni study from inside the sector’s leadership. Across the workplace survey, the national study and now three million posts, the pattern is identical. The racism is pervasive, patterned and predictable. What is scarce is not evidence. It is recognition.
The machines, briefly
This is the whole of what I want to say about AI. A general-purpose classifier learns what counts as hate from the aggregated judgement of the society that wrote its training data, which is why the researchers had to build a second model, trained on the judgements of Indigenous community annotators, because, in their words, deciding what counts as racism “cannot be separated from the experiences of the people being targeted”. A machine trained on the archive of not-seeing reproduces the not-seeing at industrial scale, and there is no technical patch for that, because the deficiency is not technical. If the country cannot see us, neither can anything it builds. The annotators were not adding sensitivity to an instrument. They were supplying the recognition the archive withholds – and who governs that judgement from here matters as much as who supplied it.
What the evidence asks
The parliamentary inquiry into racism, hate and violence directed at Aboriginal and Torres Strait Islander people now holds this study, alongside four decades of findings it corroborates. Its question – what is driving the racism – has an answer the country will not enjoy. Not the platforms, which amplify what they did not invent. The imagination itself: old, patterned, replenished on schedule, and equipped with a blindness that renarrates its violence as our dysfunction. And buried in the study’s method is the concession the sector has resisted for decades that defining racism belongs, in the first instance, to the people it targets.
Can Australia see us? Not on its own. That is what this week proves, in finding and in method alike. This is why I have spent a career arguing for rights TO, IN and THROUGH education: we were excluded from education because of race, and the imagination that did the excluding is the one this study has just weighed. What the country does with the evidence will tell us whether it wants to.
References and further reading
This Substack – the racism thread
Anderson, P. J. (2026, March 7). Is it wilful ignorance? Australia’s oldest habit. Substack.
Anderson, P. J. (2026, April 26). On the machinery of denial: And the inquiry that asks us to break it. Substack.
Anderson, P. J. (2026, May 18). Why is it so hard to think of Aboriginal people as successful? The 2026 Budget and the comfort that gets in the way. Substack.
Anderson, P. J. (2026, July 4). My culture is not the load: Fifty years of NAIDOC and a word I am retiring. Substack.
This Substack – the machine thread
Anderson, P. J. (2026, March 21). Critique is not enough. Substack.
Anderson, P. J. (2026, June 14). Who decides what the machine learns? Substack.
Anderson, P. J. (2026, August 1). The machine cannot refuse. Substack.
Articles
Anderson, P. J. (2026, February 26). Racism: Those of us with institutional power must now respond with honesty not defensiveness. EduResearch Matters (AARE). https://blog.aare.edu.au/2026/02/26/racism-those-of-us-with-institutional-power-must-now-respond-with-honesty-not-defensiveness/
The Conversation
Anderson, P. J. [et al.] (2022, February). Racism at work: A call to anti-racist action for Australian organisations. The Conversation. https://theconversation.com/racism-at-work-a-call-to-anti-racist-action-for-australian-organisations-179758
Vergani, M., & Allison, F. (2026, August 7). What three million online posts reveal about anti-First Nations hate in Australia. The Conversation.
Academic texts
Holt, L. (2020). Indigenous higher education in historical context in Australia. In P. J. Anderson, K. Maeda, Z. M. Diamond, & C. Sato (Eds.), Post-imperial perspectives on Indigenous education: Lessons from Japan and Australia (pp. 122–139). Routledge.
Moreton-Robinson, A. (2004). The possessive logic of patriarchal white sovereignty: The High Court and the Yorta Yorta decision. Borderlands e-Journal, 3(2).
Moreton-Robinson, A. (2009). Imagining the good Indigenous citizen: Race war and the pathology of patriarchal white sovereignty. Cultural Studies Review, 15(2), 61–79.
Shirodkar, S. (2019). Bias against Indigenous Australians: Implicit association test results for Australia. Journal of Australian Indigenous Issues, 22(3–4).
Reports
Australian Human Rights Commission. (2026). Respect at Uni: Study into antisemitism, Islamophobia, racism and the experience of First Nations people. Australian Human Rights Commission.
Diversity Council Australia. (2022). Racism at work: How organisations can stand up to and end workplace racism. Diversity Council Australia.
This states it well, many go to a great effort to imagine reasons not to see you. I felt the same with NO. Great effort was put into imagining spurious reasons that superficially “had nothing to do with racism”, giving people cover for their self-image as “not a racist”. That many can’t acknowledge, celebrate and be energised by First Nation people and culture is indicative of cultural bankruptcy. There’s more to life than football and barbies.
Yes indeed, the “I’m not a racist but I’ll be voting NO” attitude was quite skilfully encouraged by Dutton, Price and Rupert’s minions. It was framed almost as an act of a free-thinking individual, rather than the pull of a lack of racial and cultural imagining.
As an Australian citizen who is also a proud Welshman, i feel so proud of our First Nations people for their culture, humour, patience and connections to Country.