Abstract
Information does not become public knowledge simply because it has been published. Between an event and the account through which people encounter it lie decisions about what deserves attention, which sources are credible, how a claim should be worded, which image should represent it, and whether a record will remain accessible. Those decisions were never exclusive to journalists, but digital platforms and generative artificial intelligence have made the process more dispersed and often more difficult to inspect. This paper argues that contemporary media power is best understood not as control over individual messages alone, but as control over the transformations that occur between sources and audiences. Drawing on gatekeeping theory, agenda-setting, framing, platform studies, photographic ethics, and scholarship on archives, it develops an analytical framework for examining six connected processes: selection, framing, attribution, relations among claims, circulation, and preservation. It then proposes how THE EDITORIAL, an independent critical journalism magazine, could investigate those processes through transparent, document-based reporting. The argument is conceptual rather than a claim to findings from original fieldwork. Its central proposition is that accountable media criticism must follow information back to its source and make visible the choices that shaped its journey.
Keywords: gatekeeping; platformization; framing; generative AI; visual culture; archives; media accountability
Introduction: An event is not yet a story
A statement is issued. A journalist chooses one sentence for the headline. An editor decides where the story belongs. A platform places the article beside other material and recommends it to some users but not others. A chatbot condenses several accounts into a short answer. Months later, a researcher searches for the event and finds an archived version without the original caption, an updated headline, or the document on which the reporting relied. Each account may refer to the same event, but none performs exactly the same work.
The familiar question of whether the information is true matters, but it is incomplete. An account can be factually defensible while still altering how an event is understood. A summary may preserve a statistic but remove who produced it. A headline may accurately describe one aspect of a report while displacing its uncertainty. A photograph may be genuine but circulate with a caption that changes its political meaning. An archive may preserve a document but classify it in a way that makes it difficult to find. What is changed is not always a fact; sometimes it is the relationship between facts, sources, context, and the people represented.
This paper asks: How does information become public knowledge, and who is able to shape its meaning as it moves between institutions, media technologies, and audiences? It argues that media power operates through a chain of transformations. The key question is not only who speaks, but who selects the speech, rearranges it, attaches authority to it, makes it visible, and decides how it will be remembered. This argument does not require us to believe that editorial judgment is inherently deceptive or that technologies necessarily damage journalism. Selection is unavoidable; no newspaper, search engine, archive, or reader can attend to everything. The democratic problem arises when consequential choices are hidden, impossible to challenge, or presented as if no choice was made.
The proposed publication THE EDITORIAL takes this problem as its organizing concern. Rather than publishing general complaints that media are biased, it would investigate particular transformations: an altered headline, an AI-generated answer without a reliable source trail, a conflict photograph detached from its caption, a government statement repeated as fact, or a newspaper collection absent from a searchable catalogue. The point is to replace suspicion with demonstration. Criticism becomes stronger when it can show precisely what changed, at which stage, with what justification, and with what consequences.
1. Gatekeeping did not disappear; it multiplied
The classic account of journalistic gatekeeping begins with scarcity. A newsroom cannot report every event, and a printed page or broadcast bulletin cannot contain every available story. Shoemaker and Vos (2009) explain gatekeeping as a process in which countless possible messages are filtered into a comparatively small number that reach an audience. These choices are shaped by professional routines, organizational priorities, institutional pressures, and social conditions. Even before platforms became central to news distribution, information did not arrive in public life unmediated.
McCombs and Shaw's (1972) account of agenda-setting adds another dimension: the prominence given to issues can influence the importance audiences attach to them. Entman (1993) further clarifies that framing involves selecting aspects of perceived reality and making them salient in ways that can define problems, suggest causes, make moral judgments, or imply remedies. These theories describe related but distinct operations. Gatekeeping helps explain which stories are carried forward; agenda-setting examines the attention they receive; framing asks how the issues within them are made intelligible. An analysis that calls all three simply “bias” loses the ability to identify what a media institution actually did.
In digital news, these operations continue beyond the newsroom. A story can be carefully verified and published on a newspaper's website, yet its audience may first encounter it through a creator's video, an Instagram repost, a search result, or a recommendation feed. The Reuters Institute's Digital News Report 2026, based on surveys across 48 markets, reported that 54% of respondents used social media and video networks for news in the previous week, compared with 51% using news organizations' own websites and apps (Egan, 2026). These are survey averages across the studied markets, not a census of the entire world. Still, the comparison captures an important change in where news is encountered: publishers increasingly share control over discovery and presentation with outside systems.
It would be too simple to say that algorithms have replaced editors. News organizations still decide what to investigate, which claims meet publication standards, and what counts as a correction. Platforms introduce additional decisions about ranking, visibility, format, and audience targeting. Users also become selectors when they share, comment on, reinterpret, or refuse a story. Authority is therefore not eliminated but distributed unevenly across multiple stages. A single event may pass through several gates, and the result at one stage can determine the possibilities available at the next.
This creates an important distinction between access and control. More people may gain the ability to publish, and audiences may encounter perspectives excluded by established newsrooms. Yet broader publication access does not necessarily mean equal access to attention. The question becomes who governs discovery, whose rules are visible, and who can challenge them. Van Dijck, Poell, and de Waal (2018) argue that platforms are entangled with public values and institutional responsibilities, including fairness, accessibility, and democratic accountability. A platform should not be understood merely as a neutral channel through which already finished news passes. Its design and business arrangements form part of the conditions under which news becomes public.
For THE EDITORIAL, the article provisionally titled “The Death of the Front Page” would therefore begin with a testable question rather than a nostalgic conclusion. If three newsrooms cover the same event, which versions are prominent on their homepages, which appear in users' feeds, and what changes in emphasis across those environments? Such a comparison would not establish what every reader saw, and a visible feed ranking would not explain an opaque recommendation algorithm. It could, however, document specific differences in presentation and identify where further reporting is needed. The investigation's value would lie in the evidence of redistributed gatekeeping, not in announcing that the editor has died.
2. The new intermediary: when the answer replaces the route to the source
Generative AI introduces another stage into the information chain. Traditional search often invites a user to inspect several sources; an AI assistant can instead provide a synthesized answer within the same interface. The distinction is not absolute, because search engines have long displayed summaries and direct answers, and some AI systems include citations or retrieval features. What deserves scrutiny is the changing relationship between an original publication and the answer that appears to speak on its behalf.
The Reuters Institute reported that weekly use of AI chatbots for news increased from 7% in 2025 to 10% in 2026 across its surveyed markets (Egan, 2026). This remains a minority practice rather than a universal news habit. The report also found that 42% of respondents who used chatbots for news said they always or often clicked through to original news sources. That self-reported behavior cannot be treated as measured referral traffic, but it reminds us that an AI answer does not necessarily end the information journey. For some users it may introduce a source; for others it may become the entire account.
Source attribution is a practical issue, not a hypothetical one. In a study published by the Columbia Journalism Review, the Tow Center for Digital Journalism tested eight generative search tools using 1,600 prompts built from excerpts of news articles. The researchers reported incorrect source-identification responses in more than 60% of their tests (Jaźwińska & Chandrasekar, 2025). The result describes those particular systems, prompts, publishers, and tests in February 2025; it should not be generalized to every model, configuration, or future version. Its significance is more precise: an answer can sound confident while failing to identify the published work from which a passage came.
Attribution, however, is only one layer of the problem. Consider a document that reports an association but warns that it does not establish causation. A summary might retain the association and omit the warning. Consider two institutions that disagree over the interpretation of the same findings. A synthesized paragraph might flatten their disagreement into a single apparently settled conclusion. Neither failure is adequately described by asking whether every individual noun and number is correct. What matters is whether the account preserves uncertainty, disagreement, attribution, and the relationships among claims.
This is why the proposed article “The New Gatekeepers Don't Have Faces” should not be written as a general denunciation of chatbots. Its investigation would use bounded, documented source packets and ask multiple systems carefully specified questions about the same material. The researchers would retain the source texts, prompts, model settings, output dates, and actual responses. They would then examine whether each generated claim is supported, where credit is assigned, which qualifications survive, and whether the relationships among claims change. Cases should be reviewed by more than one human researcher, with disagreements recorded instead of concealed behind a single numerical score.
Such work would need to distinguish a model's general knowledge from the contents of the packet and avoid calling every omission an error. Summaries are necessarily selective. A concise answer that leaves out a peripheral detail may be entirely appropriate; one that removes a qualification essential to interpreting a risk estimate is not equivalent. The standard should therefore be material transformation: does the omission, rephrasing, or reordering change what a reasonable reader is likely to understand about the underlying evidence? That question is demanding, partly interpretive, and more appropriate to media research than a simplistic count of hallucinations.
The positive case for AI also belongs in the analysis. Conversational systems can help readers ask follow-up questions, translate reporting, and gain an entry point to unfamiliar subjects. The issue is not whether every answer must reproduce every source in full. It is whether convenience is paired with reliable paths back to evidence and with enough context to keep a useful explanation from becoming misleading. A responsible investigation must be prepared to find genuine improvements as well as failures.
3. The photograph is evidence, but it is also an encounter
Visual journalism makes the politics of transformation especially visible. An image is often treated as if it offers immediate access to an event: the camera was there, therefore the viewer has seen what happened. Yet a photograph has boundaries. The photographer chooses a position and a moment; an editor chooses a frame and caption; a publisher chooses its scale and placement; later audiences encounter the image among new words and images. The image may remain visually unchanged while its meaning shifts dramatically.
Susan Sontag's Regarding the Pain of Others (2003) challenges any simple assumption that seeing suffering produces a single ethical or political response. Ariella Azoulay (2008) asks us to understand photography as a set of relations among the photographer, the photographed person, and the spectator, rather than as a detachable visual object. Read together, their work suggests that viewing an image is not only an act of receiving information. It involves questions of power, responsibility, and the conditions under which another person's experience becomes publicly available.
The proposed investigation “When the Camera Becomes a Weapon” should treat the title as a question, not a verdict. A photograph may help document an abuse, expose an official denial, or preserve an event that institutions would prefer to forget. The same photograph may also be stripped of its original context, circulated for shock, or reused to support claims that its maker never intended. Calling the image a weapon before tracing its history would predetermine the answer. A stronger article would reconstruct the circumstances of its making, the original caption and publication, subsequent crops or republications, and whatever consequences can be documented rather than assumed.
This reporting faces ethical limits. The ability to reproduce an image does not settle whether it should be reproduced again. The wishes and safety of photographed people, the possible identification of vulnerable subjects, and the public value of publication must be weighed alongside the need to verify the record. At times, describing a photograph's circulation and linking to its documented publication may be more responsible than displaying a graphic image repeatedly. A magazine committed to critical visibility must be willing to investigate the act of showing, including its own.
“The Photograph That Outlived Its Caption” and “What the Crop Removed” offer smaller, more falsifiable inquiries within this wider question. If an original photograph and its later versions can be authenticated, the article can show exactly what disappeared from the frame or caption. The argument then rests on an observable editorial operation, not an assumption about what every viewer felt. The central distinction is that visibility is a condition for recognition, not a guarantee of understanding or justice.
4. Archives, translation, and the power to make a record findable
The transformation of information does not stop when the news cycle ends. Records acquire new meanings as they are collected, named, classified, digitized, translated, and searched. Michel-Rolph Trouillot (1995) argues that power enters the production of history at multiple moments, including the making and preservation of sources and the construction of historical narratives. The archive, from this perspective, is not a warehouse that merely holds the past. It is part of the machinery through which particular accounts of the past become accessible and authoritative.
A catalogue title may determine whether a document appears in a search. A missing date may prevent a researcher from connecting two records. Optical character recognition may make one language searchable more effectively than another. The absence of a digitized copy does not mean an object does not exist, yet an online search can create the impression that what it cannot retrieve was never recorded. These are not all acts of censorship. Some arise from limited budgets, damaged collections, inconsistent standards, copyright, or the practical difficulties of preservation. The consequences nonetheless deserve investigation.
The proposed feature “Who Keeps Lebanon's Newspapers?” is especially suited to document-based reporting. It could select a small number of newspaper titles and years, identify relevant institutions, and test what an ordinary researcher can actually retrieve: which issues are catalogued, which are viewable, what permissions are required, what languages and search terms work, and where the record is incomplete. The magazine should not assume in advance that particular collections are missing or inaccessible. The evidence should determine whether a gap exists, its likely causes, and whether an alternative copy survives.
Translation belongs in the same conversation. In multilingual political reporting, changing a word can alter what appears to be an accusation, a legal claim, an expression of uncertainty, or a description of responsibility. “The Translation That Changed the Meaning” could compare an original recording or text with several published translations and consult qualified speakers on the plausible range of meanings. Not every disagreement proves distortion: languages rarely offer perfectly interchangeable words. The investigation becomes credible by distinguishing an unavoidable interpretive choice from an error that materially changes the statement.
This is where a magazine concerned with both news and memory can develop a distinctive voice. It can ask not only what was published yesterday but also what can be found tomorrow. Preservation is another editorial decision, and discoverability is another form of visibility.
5. Following the institutional claim back to its origin
Not all transformations occur through new technology. Institutions have long tried to make their own language the default language of public debate. A press release may be quoted without the limitations of the report it announces. A numerical claim may survive repeated citation after its original definition has disappeared. Sponsored material may adopt the design conventions of journalism while serving a commercial purpose. A correction may be visible on the publisher's site but absent from the versions that circulated most widely.
These practices call for a distinction between evidence and institutional authority. A claim is not verified merely because it appears in an official statement, a prestigious publication, or an AI-generated answer. Each may be a useful source, but each also has a position, interests, and limitations. The Society of Professional Journalists' Code of Ethics (2014) calls for verification, context, independence, and accountability, including the correction of errors. Its principles become more meaningful when treated as actions that can be examined, not virtues that a publication declares about itself.
For instance, “The Number Everyone Repeated” could trace a widely circulated statistic through news articles, social posts, reports, and the earliest verifiable dataset. The investigation would ask whether its unit of measurement, geographic scope, time period, and uncertainty changed across repetitions. “The Correction Nobody Saw” could compare the original story's distribution with the forms of correction that followed. Neither feature should assume bad faith; an honest correction is evidence of a functioning accountability process. The public-interest question is whether readers can reconstruct how the published claim changed.
A magazine asking these questions must submit to the same scrutiny. It should distinguish reporting from opinion, disclose relevant conflicts, identify where information has not been independently confirmed, and publish corrections in ways readers can find. Otherwise, the publication would make institutional accountability its subject while keeping its own editorial authority unexamined.
6. From argument to evidence: a method for THE EDITORIAL
An intellectual identity becomes credible through a repeatable reporting practice. The defining method of THE EDITORIAL should be to follow consequential information across versions and ask what each transformation made more visible, less visible, more certain, or harder to verify. The proposed research is not one universal experiment. A photograph, an archive, a news homepage, and a language-model output require different forms of evidence. They can nevertheless be investigated through a shared discipline of provenance and comparison.
The first step is to define a bounded case and a question that could be answered differently from the writer's initial expectation. A homepage study might examine three outlets over fourteen days; an archival investigation might compare access to specified issues across two institutions; an AI study might examine a fixed packet of public documents and a recorded set of outputs. The numbers are illustrative designs, not completed samples. Choices about dates, cases, and inclusion criteria must be documented before findings are interpreted, including why some material was excluded.
The second step is to preserve versions. For every claim or artifact, the reporting file should retain the earliest verifiable source available, its publication date, author or institutional origin, stable location or archived copy where lawful, and the versions through which it traveled. A rewritten headline is not proof of manipulation unless the reporter can establish both versions and the surrounding circumstances. A missing link in an AI response is not proof that the system had access to a particular original article. The evidentiary chain must be reconstructed rather than imagined.
Third, the analysis should separate different kinds of change. Selection concerns what was included or excluded. Framing concerns emphasis, problem definition, and language. Attribution concerns whose knowledge, testimony, or labor is credited. Relation concerns whether qualifications, disagreements, or causal distinctions remain intact. Circulation concerns the routes through which an account appears and the limits of what those observations can establish. Preservation concerns what remains retrievable later. These categories overlap, but keeping them distinct improves both the rigor of the reporting and the precision of its criticism.
Fourth, interpretations should be tested against plausible alternatives. A headline might change because new facts emerged. A crop might remove a distracting object rather than suppress evidence. A catalogue gap might reflect a naming convention rather than a missing collection. An AI answer might omit a passage because the user's question reasonably asked for a brief response. The researcher should seek the strongest explanation inconsistent with the preferred thesis and report what the available evidence can and cannot settle. Where judgments depend on coding or interpretation, independent reviewers should have an opportunity to disagree, and the publication should describe how those disagreements were handled.
Finally, the reader should be able to inspect the basis for the conclusion without being asked to trust the magazine's confidence alone. That might mean linking original documents, publishing annotated comparisons, showing a methods note, or describing why a source must remain confidential. Full disclosure is not always ethical: identities, private communications, and sensitive images may require protection. But withholding evidence for legitimate reasons is different from disguising the fact that evidence is absent.
The magazine's proposed recurring desks can embody this method without dividing the publication into unrelated interests. The Source would trace where claims originate. The Edit would examine what changes through language, visuals, and newsroom judgment. The System would investigate platforms, AI, and distribution. The Record would explore archives, corrections, and institutional memory. The Reader would study interpretation, attention, and trust. The larger catalogue of thirty proposed articles is useful as a commissioning pool, but publication should be driven by what reporting can establish, not by a promise to fill every category.
The three flagship concepts show how the approach could scale. “The Death of the Front Page” follows the selection and circulation of news. “When the Camera Becomes a Weapon” follows visual evidence through captioning, reuse, and ethical judgment. “The New Gatekeepers Don't Have Faces” follows claims through algorithmic retrieval and synthesis. These are not three versions of the same story. They are distinct inquiries connected by one concern: the conditions under which information acquires public authority.
7. The limits of critique, and the obligations of an independent magazine
There is a danger in building a publication around exposing hidden power. If every difference between a source and its later representation is treated as evidence of manipulation, the investigation has already decided what it intends to find. Journalism requires editing. Translation requires interpretation. Archives require classification. Summaries require omission. A serious critical magazine must identify which changes are consequential and why, not claim that an untouched account could ever exist.
The same caution applies to commercial and technological power. Platforms can concentrate control over discovery, but they can also make new kinds of reporting accessible. AI answers can obscure source trails, but they can also assist readers who need translation, explanation, or a starting point. Visual circulation can expose injustice, but repetition can harm people whose lives have become images. Archives can reproduce institutional silences, but even incomplete records may preserve material that would otherwise be lost. These tensions are not rhetorical concessions. They determine what responsible research has to measure.
Editorial independence therefore cannot mean freedom from decisions, funding questions, or accountability. THE EDITORIAL would need a visible distinction between editorial and sponsored work, clear ownership and funding disclosures, a written correction policy, standards for handling vulnerable sources and graphic material, and an explicit policy for AI-assisted research and drafting. In particular, no generated citation or unverified quotation should enter publication simply because it appeared in a plausible answer. Interviews and original investigations should be described as completed only when they have actually been conducted.
There is also a risk in speaking about people without allowing them to challenge the frame. Reports on conflict, displacement, or marginalized communities should not reduce participants to symbols illustrating an argument about the media. Where appropriate and safe, reporting should seek the perspectives of people represented in the material under examination and offer relevant institutions the opportunity to respond to evidence-based criticism. Independence is not the absence of a point of view; it is the willingness to let evidence complicate that point of view.
Conclusion: the route matters as much as the message
The central question for contemporary media criticism is no longer adequately captured by asking who controls the newspaper or whether an algorithm is biased. Public knowledge emerges through a sequence of decisions across journalism, institutions, images, platforms, automated systems, and archives. Each stage can expand access, introduce interpretation, remove context, or preserve a record. Power lies partly in the ability to make those operations appear natural or invisible.
A stronger journalism of media power would make the route itself an object of reporting. It would show the document before the headline, the caption before the crop, the disagreement before the summary, the original dataset before the repeated number, and the catalogue record behind the search result. It would also say when evidence is insufficient and when a suspected distortion turns out to have a reasonable explanation.
This is the intellectual case for THE EDITORIAL. Its contribution would not be another declaration that the media shape reality. Scholars have demonstrated forms of that influence for decades. The magazine's contribution would be more specific and more demanding: to document how information changes, determine which changes matter, and give readers the evidence to question the authority of the final account. In a media environment where answers are increasingly easy to receive and their origins increasingly difficult to trace, that work is not a secondary form of journalism. It is one way of making public knowledge more accountable.
References
Azoulay, A. (2008). The civil contract of photography. Zone Books. https://www.zonebooks.org/books/6-the-civil-contract-of-photography
Egan, J. (2026, June 16). Overview and key findings of the 2026 Digital News Report. Reuters Institute for the Study of Journalism. https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/dnr-executive-summary
Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4), 51–58. https://doi.org/10.1111/j.1460-2466.1993.tb01304.x
Jaźwińska, K., & Chandrasekar, A. (2025, March 6). AI search has a citation problem. Columbia Journalism Review. https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php
McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36(2), 176–187. https://doi.org/10.1086/267990
Shoemaker, P. J., & Vos, T. P. (2009). Gatekeeping theory. Routledge. https://www.routledge.com/Gatekeeping-Theory/Shoemaker-Vos/p/book/9780203931653
Society of Professional Journalists. (2014). SPJ code of ethics. https://www.spj.org/spj-code-of-ethics/
Sontag, S. (2003). Regarding the pain of others. Farrar, Straus and Giroux.
Trouillot, M.-R. (1995). Silencing the past: Power and the production of history. Beacon Press. https://www.beacon.org/Silencing-the-Past-P1109.aspx
van Dijck, J., Poell, T., & de Waal, M. (2018). The platform society: Public values in a connective world. Oxford University Press. https://academic.oup.com/book/12378
Return to contents