AI Slop as a Weapon: How Towns Lose Trust in Local News

June 21, 2026

In the first week of May 2026, a row erupted on the western edge of Sydney that, on the surface, looked like the kind of parochial squabble local government produces by the cartload. Several councillors at Hawkesbury City Council, in New South Wales, dismissed the reporting of their local newspaper, the Hawkesbury Gazette, as “AI slop”. The paper had been scrutinising the council's handling of the Richmond Swimming Centre project, its mayoral minutes and a string of governance decisions. The councillors did not engage with the substance. They reached, instead, for a phrase that two years earlier would have meant nothing to anyone in the chamber: that the journalism was machine-made filler, a synthetic imitation of reporting rather than the genuine article.

The Gazette denied it, and the denial was not hard to credit. The stories were tied to council documents, named figures and verifiable financial detail. There was no evidence the coverage fitted the commonly understood definition of AI slop: the repetitive, low-effort, frequently inaccurate content that large language models now extrude across the open web at industrial scale. But the accusation did its work anyway. It reframed accountability journalism as a quality-control problem. And it landed inside a wider confrontation, because by 28 April the council's acting general manager, Will Barton, and the mayor, Les Sheather, had already moved to bar Gazette and Hawkesbury Radio representatives from council premises and meetings, citing health and safety concerns. By World Press Freedom Day on 3 May, the standoff had reached the floor of the NSW Legislative Council.

What happened in the Hawkesbury is a small story with an outsized lesson. It is the first widely documented instance of a phrase coined to describe a technological problem being deployed as a political weapon against the people whose job is to hold power to account. And it is a preview of a civic failure mode that is now arriving, simultaneously and from several directions, in towns that can least afford it.

The Phrase That Became a Cudgel

“AI slop” entered the vernacular as a description of a genuine and worsening pollution problem. As generative models became cheap and fast, the web filled with content that has the shape of writing but none of the labour: articles assembled from scraped material, padded with confident error, illustrated with images whose subjects have the wrong number of fingers. The term was useful precisely because the problem was real. Readers needed a word for the sludge.

The trouble with a useful insult is that it can be pointed in any direction. To call something AI slop is to make a claim about its provenance, that no human reporter did the work, that no editorial judgement shaped it, that it is filler dressed as fact. When that claim is true, it is a public service. When it is false but plausible, it becomes one of the most efficient instruments for discrediting inconvenient reporting ever handed to a politician. You do not need to rebut a single fact. You need only to gesture at a category and let the audience's well-earned suspicion of synthetic content do the rest.

This is the precise manoeuvre that worried observers of the Hawkesbury dispute. Local government scrutiny is exactly the sort of work AI cannot do: it requires sitting in the room, reading the budget annexes, noticing what was left off the agenda, and knowing which councillor changed their vote. To brand that work “slop” is to invert the relationship between the technology and the threat. The danger to the Hawkesbury was never that a machine wrote the Gazette's stories. The danger was that a useful word for machine-made content could be repurposed to delegitimise the human-made kind, and that enough residents, primed by genuine exposure to synthetic rubbish elsewhere, might believe it.

The councillors named in the dispute, Kotlash, McMahon and Wheeler, were not engaging in some novel theory of media criticism. They were doing what political actors have always done when reporting stings, which is to attack the messenger. The novelty is the form the attack now takes. Where a previous generation might have alleged bias or sloppiness, the contemporary version alleges inauthenticity at the level of authorship. It is an accusation perfectly tuned to a moment in which the public has every reason to doubt that what it reads was written by a person at all.

A News-Shaped Object in Colorado

To understand why the accusation is so corrosive, it helps to look at a place where the synthetic thing is real, and where a thoughtful person built it on purpose. In Longmont, Colorado, a media veteran named Scott Converse launched the Longmont News Network, an experiment in using AI “agents” as reporters. The agents scan public documents, meeting transcripts, budgets and records, and generate stories from what they find. Converse is no opportunist. He spent decades in media and technology, with stints at Apple and Paramount Global, and he had earlier founded the Longmont Observer, a non-profit local outlet that became the Longmont Leader. He started it because he was dissatisfied with the coverage his town was getting from the Longmont Times-Call after the paper moved its office out of Longmont.

In February 2026, the Times-Call turned its attention to Converse's new venture, and the headline it ran posed the question that now hangs over the whole field. Was the Longmont News Network journalism, or was it, as the headline put it, “a news-shaped object”? The phrase came from Robin Burke, a professor of information technology at the University of Colorado Boulder, who draws a careful distinction between news and what he calls news-shaped objects. AI-generated articles, in his account, fall into the latter category, because they miss the elements that make journalism journalism. “The fact that something wasn't discussed is as important as what was discussed,” Burke observed. “There's a narrative about what's happening in the city.” A model scanning a transcript can tell you what the council said. It cannot tell you what the council conspicuously avoided saying, because absence is not in the transcript. It is in the head of a reporter who has been watching for years.

The Longmont experiment has not been clean. Since increasing its publishing frequency, the platform has produced articles containing fabricated information, misspelled names, and AI-generated images that some residents mistook for real photographs. Converse, for his part, has been candid about the stakes and disarmingly modest about the result. “I don't think there's a story here,” he said. “I really believed the internet was a good thing.” He is not a villain. He is a believer in technology trying to plug a hole that the market tore in his community's information supply. That is what makes Longmont the more honest mirror of the problem. It shows what happens when synthetic local news is produced sincerely, by someone who cares, and still cannot reliably do the thing that matters.

Put the two cases side by side and the shape of the crisis comes into focus. In Longmont, a real news-shaped object is offered as a substitute for journalism, with mixed and sometimes misleading results. In the Hawkesbury, real journalism is accused of being a news-shaped object in order to discredit it. The same conceptual confusion, the inability to tell the authentic from the synthetic, powers both. And once a community loses the ability to make that distinction reliably, it becomes vulnerable to attack from either end: it can be fed filler it mistakes for reporting, and it can be persuaded that reporting is filler.

What a Town Loses When the Newsroom Goes Dark

The reason any of this matters is that local journalism does a job no national outlet, and no algorithm, has shown it can replicate. It reports on planning decisions, school budgets, the conduct of councillors, and the specifics of place that determine whether ordinary people have any visibility over the decisions that shape their daily lives. Strip that away and the consequences are not abstract. They are measurable, and a growing body of research has now measured them.

The Medill State of Local News report, the long-running census of American local journalism begun in 2015 by Penelope Muse Abernathy, a former executive at the Wall Street Journal and the New York Times, found that by 2025 the United States had lost nearly 3,500 newspapers over two decades, along with more than 270,000 newspaper jobs. In the year to the 2025 report, 136 papers closed, a rate of more than two a week. The number of news desert counties, places with no reliable local news source at all, rose to 213, and roughly 50 million Americans now have limited or no access to local news. The new digital outlets that have launched, more than 300 over five years, are concentrated almost entirely in metropolitan areas, leaving rural communities to go dark.

What happens in those places is the subject of a separate strand of Medill research, and the findings are quietly devastating. In February 2026, the Local News Initiative published survey work led by Zach Metzger, director of the Medill State of Local News Project, drawing on 1,000 respondents, half in news deserts and half in news-rich areas, polled in the summer of 2025. In news desert counties, 51 per cent of people who consume news daily get their local information from non-journalistic sources: social media groups, influencers, and friends and family. More residents leaned on these channels than on any news organisation. Forty-two per cent used social media news groups daily, 33 per cent relied on friends and family, and 30 per cent followed social media influencers. Trust in the news media sat at 46 per cent in news deserts against 59 per cent in news-rich areas. Only 10 per cent of people in news deserts had spoken to a journalist in five years, against 20 per cent elsewhere.

The most unsettling figure was not a number at all but an observation. Despite all this, around 90 per cent of people in news deserts said reliable local news was easy to access. They did not feel deprived. As Metzger put it, “You might feel like you're part of a close-knit community that knows what's going on, but places with a lack of journalism are missing an external source.” The information starvation is invisible to the starving. Tim Franklin, a Medill professor and the John M. Mutz Chair in Local News, described the diet that replaces journalism as “unvetted, un-fact-checked information bouncing around” on social platforms. Mackenzie Warren, interim executive director of the Local News Initiative, framed the deepest worry as a question of whether consumers even “value or miss what we think is so valuable”.

This is the substrate on which the AI crisis lands. A community that no longer has a newsroom does not experience the loss as a loss. It experiences a feed that feels complete. And a feed that feels complete is the ideal environment for synthetic content to take root, because there is no longer an authoritative source against which to check it. The Poynter and Medill work documents the vacuum. The next two stories show what rushes in to fill it.

The Fake Council in Yorkshire

In January 2026, the BBC's Yorkshire political editor, James Vincent, reported on what AI misinformation looks like when it targets local democracy directly. Posts began circulating that claimed to come from the City of York Council. One purported to be a council advertisement asking residents to house asylum seekers. Another sought volunteers to help take down St George's flags. A third encouraged the public to fill in potholes themselves. None of it was real.

When Vincent and colleagues at BBC Verify examined the posts, the tells were there for anyone trained to look: a council logo that was blurry and lacked detail, inconsistent fonts, spelling mistakes, and the telltale distortions in hands that betray AI-generated images. But the people sharing the posts were not trained to look, and the reach was substantial. The fake asylum seeker image had been used on accounts with more than half a million followers. The council tried to correct the record and asked the creators to retract the false material. Some refused, because the posts were earning them money. Officials voiced alarm not just about accuracy but about social cohesion, noting the volume of misinformation and disinformation about asylum seekers they were being forced to counter, and the real-world safety stakes attached to it.

Consider what this requires of a healthy information system, and what its absence does. To debunk a fake council post, you need a trusted local outlet that residents already read, that can authoritatively say “the council did not post this”, and that people will believe when it does. In York, the BBC could play that role. But York is not a news desert. Now transpose the Yorkshire scenario onto one of the 213 American news desert counties, or onto a town whose only paper has just been branded “AI slop” by its own council and barred from meetings. There is no trusted intermediary. The fake post arrives in a feed that is already the resident's primary source of local information, and there is nothing to contradict it. The misinformation does not have to be good. It has to be uncontested. The collapse of local journalism does not merely remove good information; it removes the immune response to bad information.

The Yorkshire case also exposes the economics that make the problem self-sustaining. The creators who refused to take down the fakes did so because the content paid. Engagement-driven platforms reward the inflammatory and the false, while accountability journalism, expensive to produce and frequently unwelcome to its subjects, has watched its revenue base evaporate. The machine that generates the misinformation is cheap. The institution that could counter it is going bankrupt. That asymmetry is the engine of the crisis.

When the Crowd Is the Machine

If the Yorkshire fakes represent the crude end of the threat, a paper published in the journal Science in January 2026 sketched the sophisticated end, and it should worry anyone who has ever taken the temperature of local opinion from a community Facebook group. The paper, a policy forum piece whose authors include the Nobel Peace Prize laureate Maria Ressa, the cognitive scientist and AI critic Gary Marcus, the University of British Columbia computer scientist Kevin Leyton-Brown, the network scientist Nicholas Christakis, and the misinformation researcher Sander van der Linden, among a roster of more than twenty, warned of what it calls AI swarms.

Earlier generations of bots were detectable because they were dumb: they repeated themselves, posted on schedules, and could not hold a conversation. The personas the Science authors describe are different in kind. Powered by large language models and multi-agent systems, they can enter digital communities, participate in discussions, and influence viewpoints at extraordinary speed. They adapt to feedback, coordinate instantly, and maintain consistent narratives across thousands of accounts. A single operator can run a vast network of these voices, each one adopting local language and tone, each one indistinguishable from a neighbour. The systems can run millions of small experiments to learn which messages persuade, refining their approach in real time and manufacturing what looks like organic, widespread public agreement.

The civic danger here is not simply that a town might be lied to. It is that a town might be presented with a counterfeit of its own opinion. Manufactured consensus is more corrosive than a single fake post, because it hijacks the social proof that humans use to decide what is normal, safe and true. If a community forum appears to be full of locals furious about a planning application, or warmly supportive of a developer, or convinced a councillor is corrupt, residents calibrate their own views accordingly. They do not know the chorus is synthetic. Leyton-Brown drew out one of the stranger long-term consequences. “We shouldn't imagine that society will remain unchanged as these systems emerge,” he warned. “A likely result is decreased trust of unknown voices on social media, which could empower celebrities and make it harder for grassroots messages to break through.” In other words, the swarm does not only deceive; it poisons the well, teaching everyone to distrust the very strangers whose voices local democracy depends on hearing.

Now reassemble the pieces. A news desert leaves a community without a trusted source and unaware it is missing one. Into that vacuum flow fake institutional posts of the Yorkshire variety, uncontested because there is no newsroom to contest them. Layered on top, AI swarms manufacture a fake version of local sentiment that residents mistake for the real mood of their own town. And when an actual journalist does manage to report something true and inconvenient, the “AI slop” accusation, weaponised in the Hawkesbury, stands ready to discredit it. Each failure makes the others worse. The community loses not just its information but its ability to tell information from its imitation, which is the more fundamental loss, because it is the loss from which there is no easy recovery.

The Specific Civic Harm

It is worth being precise about what is actually at stake, because vague invocations of “trust” and “democracy” do not capture the mechanism. The harm is the severing of the link between citizens and the decisions made in their name.

Local journalism is not interchangeable with national coverage. A national outlet will never report that a particular council quietly rezoned a particular floodplain, or that a school's budget was reallocated away from special-needs provision, or that a contract went to a councillor's associate. Those facts are too small to register nationally and too consequential to ignore locally. They are the texture of governance at the scale where most people actually encounter the state. When the reporting of those facts disappears, or becomes indistinguishable from synthetic noise, the decisions do not stop being made. They simply stop being seen. Power that operates unseen is power that operates unchecked, and the Hawkesbury dispute is instructive precisely because it shows officials moving to make their conduct less visible, by branding the coverage fake and barring the reporters, at the very moment that coverage became inconvenient.

There is a second-order harm that compounds the first. The “liar's dividend”, a term that long predates the current AI wave, describes the benefit that accrues to bad actors once the public knows that fakery is possible. If anything can be fabricated, then anything inconvenient can be dismissed as a fabrication. The Hawkesbury accusation is the liar's dividend applied to journalism itself. Once a community accepts that AI slop exists, and it does, the door opens to dismissing genuine reporting as slop whenever it stings. The very real problem of synthetic content provides cover for the very old problem of evading accountability. The technology supplies the alibi; the politics supplies the motive.

The third harm is the most insidious, and it is the one the Medill survey captured. It is the disappearance of the felt need for journalism at all. A population that gets its civic information from feeds, influencers and gossip, and that reports finding reliable local news “easy to access” while living in a documented news desert, has lost not only the supply but the demand. You cannot organise a campaign to save something you do not know you have lost. This is why the crisis is so resistant to market solutions. The market signal that would normally summon a replacement, consumer demand, has itself been anaesthetised.

Who Can Actually Prevent It

The temptation, faced with a problem this distributed, is to reach for the largest available lever and demand that someone pull it. But there is no single lever, and the actors capable of pulling the various smaller ones are scattered across very different domains. Prevention, if it comes, will be a matter of several parties doing their separate jobs, and the honest assessment is that some are better placed than others.

The platforms sit closest to the technical reality and have done the least with that proximity. The Yorkshire fakes spread because the platforms that hosted them rewarded engagement over accuracy and paid the creators who refused to take the fakes down. The AI swarms described in Science are a platform-level problem by definition, because they live inside the social graphs that platforms own and could, in principle, instrument. Robust provenance standards, the cryptographic labelling of authentic institutional accounts, the rapid de-amplification of content impersonating public bodies, and the genuine detection of coordinated inauthentic behaviour are all within the technical reach of the largest companies on earth. The obstacle has never been capability. It has been the absence of any incentive strong enough to override the business model, which is exactly the gap that regulation exists to fill.

Regulators and lawmakers hold the instruments that can change those incentives, and a few are beginning to use them. The NSW response to the Hawkesbury ban is a small but real example of institutional friction working as intended. John Ruddick, a member of the Legislative Council, lodged a motion condemning the exclusion of the Gazette and Hawkesbury Radio, calling it, in characteristically blunt terms, “outright fascism displayed by Hawkesbury City Council”. The state's Local Government Minister, Ron Hoenig, requested an investigation by the Office of Local Government, and SafeWork NSW examined the safety justification the council had offered. None of this addresses synthetic content directly. But it demonstrates the principle that matters most: that the right of accountability journalists to be in the room is not the council's to revoke, and that the “AI slop” framing does not survive contact with a functioning oversight system. The deeper regulatory task, mandatory provenance and disclosure for synthetic content, liability for platforms that profit from impersonation, and protections for journalists' access, remains largely unbuilt.

The newsrooms themselves are not passive in this, and the Hawkesbury Gazette offered a small masterclass in how an outlet holds the line. Rather than litigate the “AI slop” smear in the abstract, the paper anchored every disputed story to council documents and public statements, making provenance its defence. Its publisher, Kooryn Sheaves, vowed to keep covering meetings “from the footpath, if necessary”, reporting “during evening meetings, in the dark, with a head torch and a thermos of hot tea”. That is more than defiance. It is the recognition that in an environment of synthetic doubt, a journalist's most valuable asset is demonstrable, checkable, human provenance: the visible fact of having been there. Transparent sourcing, clear bylines, published methods and, increasingly, cryptographic content credentials are becoming not optional extras but the working definition of trustworthy local reporting.

Funders and the public hold the levers the market has dropped. The Medill research is supported by the MacArthur Foundation, and the more than 300 digital startups launched over five years show that philanthropic and community models can stand up real reporting where advertising no longer will. But those startups cluster in cities, and the rural news deserts that are most exposed to synthetic capture are the least served by them. Closing that gap is a deliberate choice that funders, and the communities themselves, would have to make. Which returns the question to the residents, who are simultaneously the victims of the crisis and, uncomfortably, the only constituency with the standing to demand the rest of it be fixed. The Medill finding that they do not feel the loss is the single hardest obstacle to clear, because every other intervention depends on a public that knows what it is missing and is willing to pay, in attention or money or votes, to get it back.

The Distinction Worth Defending

The thread running through Hawkesbury, Longmont, Yorkshire and the Science paper is a single, deceptively simple capacity that is now under sustained assault: the ability of an ordinary person to tell authentic reporting from its machine-made imitation. Scott Converse's news-shaped object and the Hawkesbury councillors' “AI slop” jibe are two sides of one coin. Both depend on, and both deepen, the public's growing inability to make that distinction with confidence. The fake York council posts and the AI swarms exploit the same confusion from the other direction, flooding the zone with the synthetic until the genuine can no longer be picked out.

Robin Burke's formulation is the one to hold onto, because it names what is actually at risk. The value of journalism was never only the information it conveyed. It was the judgement embedded in the choosing: the knowledge of what was left unsaid, the narrative of what is happening in the city, the reporter who notices the agenda item that vanished and asks why. A model can produce text that looks like that. It cannot, yet, produce the judgement, and it certainly cannot sit in a council chamber for a decade and develop the institutional memory that makes the judgement worth having. The civic harm is what happens when communities forget there is a difference, and the people who could remind them are either disappearing for want of funding or being told, by the very officials they cover, that they were never real to begin with.

The Hawkesbury Gazette is still reporting, from the footpath if it has to. That it has to is the warning. The question of who can prevent the wider harm has an unsatisfying but honest answer: everyone with a relevant lever, acting at once, before the communities at greatest risk lose not just their newsrooms but the memory of why a newsroom mattered. The places already in the dark are the ones who will not raise the alarm, because they no longer know the lights have gone out.

References and Sources

Hawkesbury Gazette. “Councillors label Gazette reporting 'AI slop'.” 8 May 2026. https://www.hawkesburygazette.com/councillors-label-gazette-reporting-ai-slop/
Hawkesbury Gazette. “NSW Parliament motion condemns Hawkesbury media ban as pressure mounts on Council.” May 2026. https://www.hawkesburygazette.com/nsw-parliament-motion-condemns-hawkesbury-media-ban-as-pressure-mounts-on-council/
Hawkesbury Gazette. “Council Bans Gazette from Meetings Citing Safety Concerns.” 2026. https://www.hawkesburygazette.com/council-bans-gazette-from-meetings-citing-safety-concerns/
Hawkesbury City Council. “Statement – exclusion of Hawkesbury Gazette and Hawkesbury Radio.” May 2026. https://www.hawkesbury.nsw.gov.au/_resources/media-releases/2026/may/statement-exclusion-of-hawkesbury-gazette-and-hawkesbury-radio
Lyle, London. “Longmont media veteran launches AI news site, but is it just 'a news-shaped object'?” Daily Times-Call, 8 February 2026. https://www.yahoo.com/news/articles/longmont-media-veteran-launches-ai-153200538.html
MediaPost. “Around the Net In Media: Longmont News Network Pursues AI-Based News.” 9 February 2026. https://www.mediapost.com/publications/article/412643/longmont-news-network-pursues-ai-based-news.html
Poynter. “When local news disappears, people turn to social media feeds, influencers and gossip.” February 2026. https://www.poynter.org/business-work/2026/where-do-people-in-news-deserts-get-information/
Local News Initiative, Northwestern University Medill School. “With no local news, those in news deserts turn to social media feeds, influencers and gossip.” 10 February 2026. https://localnewsinitiative.northwestern.edu/posts/2026/02/10/news-deserts-social-media-local-news-medill-survey/index.html
Local News Initiative, Northwestern University Medill School. “The State of Local News 2025.” October 2025. https://localnewsinitiative.northwestern.edu/projects/state-of-local-news/2025/
Medill School, Northwestern University. “News deserts hit new high and 50 million have limited access to local news, study finds.” October 2025. https://www.medill.northwestern.edu/news/2025/news-deserts-hit-new-high-and-50-million-have-limited-access-to-local-news-study-finds.html
Vincent, James. “How AI is posing a threat to democracy in Yorkshire.” BBC, January 2026 (republished via Yahoo News). https://www.aol.com/articles/ai-posing-threat-democracy-yorkshire-080156882.html
TechRepublic. “Fake UK Council Posts Show the Power of AI Misinformation.” January 2026. https://www.techrepublic.com/article/news-uk-council-ai-misinformation/
ScienceDaily. “AI swarms could hijack democracy without anyone noticing.” 20 April 2026. https://www.sciencedaily.com/releases/2026/04/260420014748.htm
Schroeder, D.T., Cha, M., Baronchelli, A., Bostrom, N., Christakis, N.A., Garcia, D., Goldenberg, A., Kyrychenko, Y., Leyton-Brown, K., Lutz, N., Marcus, G., Menczer, F., Pennycook, G., Rand, D.G., Ressa, M., Schweitzer, F., Song, D., Summerfield, C., Tang, A., Van Bavel, J.J., van der Linden, S., and Kunst, J.R. Policy Forum, Science, 22 January 2026; 391 (6783): 354.
University of British Columbia. “AI swarms could hijack democracy, without anyone noticing.” 2026. https://news.ubc.ca/2026/01/ai-swarms-could-hijack-democracy-without-anyone-noticing/
Poynter. “An alarming number of independent publishers and small chains closed papers last year, new Medill study finds.” 2025. https://www.poynter.org/business-work/2025/medill-report-local-news-closures-independent-papers-news-deserts/
Nieman Journalism Lab. “In Medill's latest State of Local News report, a 'festering, 20-year-old problem' looms larger than ever.” October 2025. https://www.niemanlab.org/2025/10/in-medills-latest-state-of-local-news-report-a-festering-20-year-old-problem-looms-larger-than-ever/

Tim Green UK-based Systems Theorist & Independent Technology Writer

Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.

His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.

ORCID: 0009-0002-0156-9795 Email: tim@smarterarticles.co.uk

Listen to the free weekly SmarterArticles Podcast

Discuss...