Everyone Sees Something Important, Just Not the Same Thing: Agenda-Setting After the End of the Shared Front Page

At 8.10 on Monday morning, two neighbours open their phones. One sees a transport strike, a parliamentary inquiry and flo oding in a nearby district. The other sees a celebrity interview, a market rally and a furious debate about a university several hundred miles away. Both feeds contain real journalism. Neither person is necessarily trapped in falsehood.

At the lift, they have nothing to argue about.

That sounds peaceful until the flooded district needs volunteers, the council closes a consultation and the strike changes how thousands of people reach work. Democracy does not require identical opinions. It does require some common objects of attention: things about which disagreement can become action.

The old front page was never a neutral mirror. It excluded, simplified and concentrated editorial power. But it made its hierarchy visible. A personalised feed creates a front page for each person — and usually hides the fact that anybody else received a different world.

Agenda-setting was never about mind control

The classic agenda-setting study by Maxwell McCombs and Donald Shaw compared the issues emphasised by news media during the 1968 US presidential campaign with the issues undecided voters considered important. Its central proposition was carefully limited: media are particularly influential in shaping the salience of issues — what people think about — rather than mechanically determining what people think (McCombs & Shaw, 1972, pp. 176–187).

Salience is not agreement. A conservative and a socialist can place housing at the top of the agenda for opposite reasons. Their conflict may be fierce, but it is politically usable because it has an object. Budgets can be examined, claims tested and representatives questioned.

Agenda-setting also does not mean that audiences are empty containers. People bring experience, identity and conversation to their media use. Different outlets compete, events can force themselves onto the agenda and publics can organise counter-agendas. Online public opinion changes the relationships among media, networked actors, visibility and participation rather than producing a simple one-way effect (Nowak-Teter, 2023, pp. 141–156, 191–214).

The concept remains powerful because attention is scarce. An issue that receives repeated, prominent and timely coverage becomes easier to recall and more available as a measure of public importance. An issue that is absent does not cease to exist; it becomes harder to make collectively consequential.

The shared front page was flawed — and still did a job

It is tempting to turn the newspaper front page into a lost civic fireplace around which an entire nation once sat sensibly. History was less charming. Editors were disproportionately drawn from dominant groups. Local, minority and working-class experiences were often marginal. A small number of institutions possessed enormous gatekeeping power.

Yet the printed hierarchy had three civic properties worth noticing.

First, it was inspectable. Readers could see which story was first, which photograph was largest and what had been omitted. Critics could compare the choices (Napoli, 2019, chs. 4–6).

Second, it was shared enough to be contested. Readers did not receive identical media diets, but large groups encountered overlapping priorities. A union, opposition party or community organisation could challenge the same visible agenda.

Third, it was editorially attributable. A masthead, editor and set of professional norms stood behind the selection. Accountability was imperfect, but it had an address.

Personalised distribution weakens all three. Ranking may differ for every user, the comparison page is unavailable and responsibility is distributed among newsroom, platform, model, data and behaviour. The selection still happens. It simply feels like weather.

From one hierarchy to millions of private salience maps

Personalisation is useful: location, accessibility preferences and specialist interests can improve a feed. The problem is not relevance. It is the disappearance of the boundary between personal relevance and public importance.

Three feedback loops drive that disappearance.

The first is behavioural. A person pauses on an outraging story, so the system learns that outrage is relevant. It shows more of the topic, making it more familiar and apparently more important. Behaviour becomes input; ranking becomes experience; experience produces more behaviour.

The second is commercial. Platforms often optimise measurable engagement, retention or conversion. A slow-moving investigation into procurement may lose to a conflict with a face, a villain and a fresh notification. This is not because an algorithm hates municipal accountability.

Hate would at least be a position. The system is indifferent in a direction set by incentives.

The third is identity-based. Users are grouped through inferred interests, location, demographics and affinities. An issue is shown because ‘people like you’ responded before. The feed gradually teaches the user what people like them should care about (Pariser, 2011).

One risk is that “Filter bubbles restrict the diversity of perspectives and deepen social polarisation” (Łosiewicz, 2025, p. 106; author’s translation). Agenda fragmentation goes one step earlier. People may lose not only alternative perspectives but awareness that the issue belongs to the public agenda at all.

Generative answers make the agenda even less visible

A feed still displays a sequence of items. A generative news assistant can replace the sequence with a single answer. Ask, ‘What matters today?’ and the system selects, ranks, compresses and narrates without showing the rejected candidates.

This creates agenda-setting by synthesis. The answer may be accurate sentence by sentence while its architecture remains opaque. Which geography counted as ‘today’? Which sources were retrieved? Was public importance inferred from newsroom prominence, search popularity, platform engagement or the model’s own patterns? Did a developing story disappear because evidence was sparse, or because available sources were in a less represented language?

Conversational delivery adds authority. A list of headlines looks selected; a fluent paragraph can feel like knowledge itself. Follow-up questions may deepen the chosen path while leaving the initial omission untouched. The user asks why the inquiry matters, then who testified, then what happens next. Nobody asks about the flood that never entered the first answer.

This does not prove that generative interfaces always narrow agendas. They could deliberately introduce diversity, local relevance and under-covered

issues. But the newsroom must treat retrieval and omission as editorial decisions, even when a supplier calls them model behaviour.

The cost of a public that cannot see its own priorities

The person who pays may be a resident whose issue never crosses the threshold of personalised relevance. A care-home closure affects a small group intensely but attracts little broad engagement. A local journalist covers it, yet a distribution system predicts limited interest and suppresses the story. The people most affected see it; everybody whose pressure could change the decision does not.

The cost may also fall on the reader. Personalisation promises respect for individual choice, but choice made from an invisible menu is thin autonomy. A person cannot decide to attend to a matter they were never told existed. Freedom of information is not only the freedom to select among offered items; it includes enough orientation to understand what the wider community is confronting.

Journalists pay too. If performance is measured only at personalised item level, reporters receive a distorted lesson about value. They may learn that a story ‘failed’ when it was never distributed beyond a predicted niche. News judgement begins to follow the ranking system’s map of likely attention.

Finally, democratic institutions lose a synchronisation mechanism. Public action often depends on timing: a consultation closes, a vote takes place, a safety warning expires. A fragmented public may discover an issue eventually and still arrive after the decision. In politics, lateness is sometimes a form of exclusion.

Putting people at the centre therefore requires more than maximising satisfaction. It means protecting dignity — including the dignity of groups whose problems are not popular — and practical agency: the ability to know, object, organise and demand an answer.

Two layers: personal pathway, public core

A responsible news service need not choose between one national front page and total personalisation. It can separate two functions.

The personal pathway adapts format, order and depth to a reader’s needs. It can foreground location, accessibility preferences, followed topics and available time. It should offer meaningful controls and explain important inferences.

The public core is a limited, visible set of matters the newsroom judges necessary for common orientation. Every user can find it; critical items cannot be silently removed by behavioural personalisation. The core need not occupy the same screen position for everyone, but it must remain recognisable and accessible.

This is not a daily catechism. A public core should be plural, revisable and small enough to mean something. It may include urgent safety information, major exercises of public power, decisions with broad or severe consequences and subjects systematically under-covered despite significant impact.

The newsroom’s authority to define that core must itself be open to challenge. Otherwise the shared agenda merely returns wearing a digital badge.

The Public Agenda Register

The Public Agenda Register is a visible record of why a topic enters, remains in or leaves the public core. It turns hierarchy from a hidden output into an accountable editorial practice.

1. Name the issue, not just the story. ‘Minister gives interview’ is an event. ‘Access to mental-health care’ is an issue. The register connects episodic updates to the longer public matter, preventing a famous speaker from becoming the agenda by gravitational force.

2. State “why now”. Record the trigger: a decision, deadline, new evidence, escalating harm or significant public dispute. Distinguish

timeliness from virality. ‘Trending’ may be evidence of attention, but not sufficient evidence of importance.

3. Publish the inclusion criteria. Note the scale, severity, reversibility, distribution of effects and public-power dimension. A small group facing irreversible harm may warrant inclusion even when raw audience numbers are low.

4. Identify affected people and missing voices. List who bears the consequences, who has been quoted and who has not yet been heard. This is not demographic box-ticking. It tells editors where knowledge is incomplete and gives communities grounds to challenge representation.

5. Link evidence and distinguish status. Show primary documents, reporting and relevant data. Mark established facts, contested claims, editorial judgement and unresolved uncertainty. In a generative interface, each summary should preserve a path back to this record.

6. Show the distribution rule. State whether the issue appears in the public core, a local core or a personalised pathway; whether alerts were used; and which accessibility formats are available. Test whether language, disability settings or low engagement change access to essential material.

7. Record omissions and dependencies. If the newsroom relies on a platform or model that does not expose ranking logic, say so. Unknown should appear as a risk, not as empty white space. Log stories excluded because evidence was insufficient and define what would reopen assessment.

8. Set a review date and route of appeal. Public importance changes. Assign an editor to reconsider the issue, publish major corrections and receive challenges from readers or affected groups. Provide a human contact; an automated form that can only classify complaints already imagined by the newsroom is not an appeal.

A monthly audit should compare the public core with actual distribution. Which topics reached most users? Which groups saw them late? Which issues were repeatedly personalised away? Which communities appeared mainly as problems rather than sources? The results should inform both editorial policy and technical settings.

What a register cannot repair

A register can create new power as well as accountability. Editors may turn criteria into a rigid formula, while organised actors learn to game severity or virality signals. Publishing too much operational detail can expose sources or make manipulation easier. Small newsrooms may lack staff for extensive documentation.

It also cannot force a dominant platform to share its agenda logic. A newsroom may build a careful public core that an external recommender then buries. Distribution contracts and platform policy therefore belong in the audit.

Nor should common orientation become compulsory sameness. Minority publics need spaces to form their own agendas, and personalisation can surface neglected expertise. The aim is overlap, not uniformity; contestable common ground, not a return to three editors deciding what a country is allowed to notice.

Media education remains essential: critical evaluation, source selection, socio-cultural context and AI literacy are connected levels of contemporary competence (Cymanow-Sosin, 2026, pp. 22–24). But telling individuals to ‘escape their bubble’ is insufficient when they cannot inspect the ranking architecture.

Tomorrow morning, the two neighbours may still open different feeds. One can receive the transport update in audio; the other can begin with market news. Yet both should be able to see that the council vote, flood and care- home closure belong to a common civic horizon — and why.

The public does not need one opinion. It needs enough of a shared map to find the place where opinions can meet.