Generative AI is making competent content easier to produce. It can assemble a clear explanation, smooth an awkward draft, propose a structure, and imitate almost any familiar style. The floor rises, but the field also fills with work that is difficult to distinguish. Publishers then face a harder problem: giving readers a reason to choose one source over another.
The answer is unlikely to be more volume. It may not even be better prose. As production becomes less scarce, a publication’s defensible value shifts toward something AI can imitate sentence by sentence but cannot establish on its own: a recognizable point of view.
Point of view is often mistaken for tone. It is treated as a layer of personality applied after the reporting is finished—a preference for sharp sentences, a degree of formality, or a recurring set of phrases. Those qualities are increasingly easy to reproduce. A real point of view is more demanding. It is the pattern readers discover in what a publication notices, which questions it asks, what evidence it respects, how it weighs competing explanations, and what it is willing to stand behind.
That pattern becomes more valuable as the supply of plausible content expands.
AI can raise quality while narrowing difference
Bad writing is only part of the risk. In many cases, AI can make writing better.
A 2024 experiment published in Science Advances assigned 293 people to write short stories either without AI assistance or with access to ideas generated by GPT-4. The AI-assisted stories were evaluated as more creative, better written, and more enjoyable, with the largest gains among participants who initially scored lower on a creativity measure.
But the researchers found a second effect: the AI-assisted stories were more similar to one another than those written without AI. Individual results improved while the collective range of results narrowed.
Short fiction is not journalism, and one experiment cannot predict the future of publishing. The tension it reveals is nevertheless important. A tool can improve the quality of each item while making the larger field feel more alike. My colleague Martin Anderson has also covered preliminary research suggesting creative outputs are converging across competing model families. If many publishers use similar systems to produce similarly competent work, competence becomes a baseline rather than a reason to choose.
Publishing once rewarded the ability to produce consistently: more stories, faster updates, cleaner copy, broader coverage. Those capabilities still matter, but they are easier to acquire than they once were. The harder advantage is deciding what all that capacity should be used to say.
Point of view is a record of decisions
A publication can sustain a point of view while allowing its articles to reach different conclusions. Reporting can remain open to disagreement without forcing every writer into a house personality. What matters is that the publication’s judgment becomes legible across time.
Readers learn which developments it treats as consequential and which it treats as noise. They learn whether it prefers novelty or proof, access or independence, speed or context. They learn how it handles uncertainty, whose expertise it considers credible, and whether it corrects itself when the evidence changes.
None of those choices lives entirely at the sentence level. They appear in commissioning, sourcing, editing, placement, follow-up, and omission. A model can be prompted to mimic the finished surface, but it cannot retroactively acquire the history of decisions that made the surface meaningful.
Style and point of view can move in opposite directions. A publication may have vivid prose but no discernible judgment; every subject receives the same dramatic treatment because the goal is attention. Another may sound restrained yet feel unmistakable because its standards and interests are consistent. Recognition comes from coherence, not decoration.
Coherence can harden into predictability, ideology, or a refusal to admit inconvenient facts. Distinctiveness is no substitute for accuracy. Editorial identity earns value by remaining answerable to evidence. It tells readers how a publication sees while leaving room for reality to change the view.
Audiences are choosing identifiable sources
Audience research cannot isolate a single reason people choose recognizable sources. It does show how important identifiable people and trusted institutions have become.
Pew Research Center found that 21% of U.S. adults regularly get news from social-media news influencers, rising to 37% among adults under 30. Among those who use these sources, 65% said influencers helped them understand current events and civic issues, while 70% said the information was at least somewhat different from news they received elsewhere.
Personality-led media is not inherently trustworthy. The same shift can reward confidence without evidence and familiarity without accountability. Even so, many readers are choosing more than a subject. They are choosing an identifiable interpreter.
Institutions offer a different form of recognition. The Reuters Institute’s 2025 Digital News Report, covering 48 markets, found that audiences across generations still value trusted brands with a history of accuracy even as engagement with traditional news sources declines. An institution and an individual are not interchangeable, but both give the reader a basis for expectation.
Google says search users are increasingly seeking original posts, unique perspectives, and firsthand voices. The company has not disclosed enough methodology to treat that claim as independent measurement. Its explanation is nevertheless revealing: generic information can often be summarized on the results page, while lived knowledge and distinctive judgment give people a reason to continue to the source.
The publication becomes the filter
For much of digital publishing, scale rewarded coverage breadth. A publication could increase its surface area by answering more queries, covering more variations of a subject, and publishing against a larger set of keywords. Generative AI extends that logic almost without limit. It can help create an article for every conceivable question.
Yet unlimited coverage does not create unlimited attention. It transfers the burden of selection to the reader, who now faces more adequate answers than anyone can meaningfully evaluate. In that environment, the publication itself becomes a filter. Its value lies partly in reducing the reader’s need to decide what deserves consideration.
That makes point of view an economic asset as well as an editorial one. A reader returns because past choices created a useful expectation about future choices. The publication has taught the audience what its attention means.
AI can support that work. It can extend research capacity, compare sources, test an argument, and help express an idea across formats. But if it is used mainly to expand output, it can leave a publication sounding increasingly polished and progressively less necessary.
Loudness and certainty are weak substitutes for recognition. A publication becomes recognizable when its choices form a coherent record: this mattered, this evidence counted, this uncertainty remained, and this conclusion was worth the reader’s time. AI can help make those choices visible. The history behind them still has to be earned. When competent content is everywhere, that record of judgment becomes the scarce asset.
