Learnings for Government Comms from Opening the Black Box Phase 2

AI summaries and LLMs are now shaping public sentiment toward government performance, and the narrative they construct bears little resemblance to the one produced by the daily news cycle. That is the central finding of Medianet's Opening the Black Box Phase 2 report, a study of how large language models – ChatGPT, Gemini and Claude – cite and represent brands and public institutions.

We analysed 957 LLM responses and hand-coded 11,127 citations, with a focus on the Victorian Government's presence in AI responses tied to the My Melbourne Road campaign. Four findings stand out for anyone working in public sector communications.

Medianet Managing Director Amrita Sidhu frames the shift as a governance issue, not a technology trend. "LLMs are a stakeholder that will need their own type of management," she said. "They are an emerging reputational channel that needs to be measured, monitored and managed. Our research is the first step in building the framework to do that."

Owned content is the largest single source LLMs cite

Government sources made up 44% of all citations in the study, ahead of corporate content at 23% and traditional media at 16%.

Individual Victorian Government domains carried the weight. vic.gov.au was cited at 5.7%. tac.vic.gov.au 3.8% , transport.vic.gov.au 2.3%, bigbuild.vic.gov.au 2.2%, premier.vic.gov.au 2.2%, and parliament.vic.gov.au 2%.

Combined, these six domains account for close to a fifth of every citation recorded in Phase 2. Victorian Government sources generated more than 3,500 citations, much more than any single corporate brand in the study.

LLMs treat structured, publicly accessible government content as fact. Agencies and portfolios that publish detailed, well-organised information about their legislative agenda are cited on the strength of it, which makes information architecture a lever for shaping public narrative, apart from being a service-delivery function.

AI models rate government performance far more favourably than the press does

One of the starkest findings of the report was the discrepancy in sentiment between traditional media coverage and LLMs.

In the data set, 50% of media mentions in traditional media were negative or unfavourable. For LLMs, only 14% of responses on the same subject matter were unfavourable, while 43% were favourable.

It is the same government in the same legislative period with two contradictory readings of public sentiment, depending on which channel is consulted.

Jacquie, our Head of Insights, traced the gap back to sourcing. "In Phase 2, we drew a connection between the sentiment towards Vic Government and the content information contained in LLM responses," she said. "These were similarly lifted from Vic Government websites."

Why the discrepancy?

The answer is in the sources.

While journalists apply scrutiny and editorial judgement to their reports, LLMs pull from what has been published and structured, and government “own” content such as official reports, ministerial websites, fact sheets and policy pages are preferred as sources by LLMs because of their authoritative nature and structure.

This “bias” towards government sources causes a difference in sentiment. The content LLMs pull from is written to inform rather than critique, and the sentiment they generate reflects that. A department or portfolio's own site is doing more reputational work in AI environments than most communications teams have accounted for.

An issue that fades from the news can still get cited in AI

News coverage of a government subject typically runs its course in two to four weeks. AI models operate on a different clock. Once an issue is indexed, it can stay in a model's responses indefinitely.

Adam Palmer, Media Analyst at Medianet, in our recent webinar, pointed to the Royal Commission into Antisemitism and Social Cohesion as a live case study. University leaders were absorbing a rough week of headlines, but he argued the headlines weren't the real risk. He said, “The important thing for the vice-chancellors and university leaders and their comms teams is that they influence how they are portrayed then. That means being very careful with the statements and written submissions they make to the Royal Commission, because that is what will influence the official record.”

The same principle holds for any government agency facing a parliamentary inquiry or review. Formal submissions and the public record carry more weight in AI environments than daily headlines do. The data backs this up. The 2019 Banking Royal Commission was still surfacing in LLM responses in 2025, six years on. A December 2024 ASIC ruling that drew heavy traditional media coverage hadn't reached LLM responses at all by the time of analysis. Some issues take six months or longer to appear in AI responses. Once they do, they don't leave.

The accountability gap

The dominance of owned content raises a harder question for public accountability. If government material drives most AI citations, and AI models soften controversy by default, does ministerial accountability suffer as a result?

A critical story about government spending or policy failure can reach the public within days through traditional media. The same story might not surface in AI responses for six months, if it surfaces at all. In that window, a constituent asking an AI assistant about the same issue gets a materially different picture of public sentiment than the one reading the news. Public sector reporting specifically, where scrutiny of spending and policy decisions is the point, the distance adds up.

It's compounded by a separate structural issue. Much of Australia's highest-quality journalism sits behind paywalls or crawl restrictions, making it largely invisible to AI models regardless of its accuracy or public interest value.

Cited, but not credited

The research also flagged a pattern worth watching. In 137 responses, AI models cited RACV content verbatim without naming RACV at all. For a campaign built around public education, that may not matter as the message still resonates, but for a government program designed to build recognition of a specific initiative, funding announcement or agency, it can be a measurement problem. The department can do the work and get none of the visibility, and the narrative around its own legislative wins goes uncredited.

What does this mean for government comms?

Three priorities that follow directly from the data.

• Treat owned content as core infrastructure. A well-structured, detailed government website is the strongest driver of favourable public sentiment in AI representation available to public sector teams. It is worth developing further.
• Your external content is getting cited the most. Parliamentary inquiries and legislative reviews all shape how AI models represent government performance for years after the headlines fade. They have always been important, but structuring them for LLM Visibility will make a difference.
• Extend monitoring well past the news cycle. An issue that has gone quiet in traditional media can still be building toward an AI citation months later, and once cited, it tends to stay cited, shaping the legislative narrative long after the headlines move on. Audit AI visibility with the same discipline applied to media coverage.

Medianet can track this via public-sector-specific intelligence that tracks narratives shaping public opinion via the media and LLMs.

These findings are drawn from Phase 2 of Opening the Black Box, our ongoing research into generative citation analysis. Read the top findings from the report, or download the full report here.

FAQ's

Do AI models rate government performance differently to traditional media?

Yes, significantly. Half of traditional media coverage of the Victorian Government was unfavourable, compared to just 14% of LLM responses on the same subject.

Why do LLMs treat government content more favourably than the news does?

LLMs pull heavily from government websites and fact sheets, which are written to inform, not critique. In Medianet's research, government sources made up 44% of all citations, more than corporate content or traditional media combined.

Can a government issue still show up in AI responses after it's dropped out of the news?

Yes. The 2019 Banking Royal Commission was still surfacing in LLM responses six years later, while a 2024 ASIC ruling with heavy media coverage hadn't reached LLM responses at all. Some issues take six months or more to appear, then stick around indefinitely.

What should government comms teams do to manage their AI visibility?

Invest in strong owned content, extend monitoring past the news cycle, treat formal submissions as long-term assets, and audit AI visibility separately from traditional media. Medianet does this through its public-sector intelligence solutions. Learn more here.