By Emre Zor
@redlineproject
AI disclosure: ChatGPT was used to outline part of this story. Perplexity was used to generate research sources. Sora was used to create images. The summary and video were generated by Summarizer 2 Custom GPT and NotebookLM, respectively.
Summary: Artificial intelligence is increasingly integrated into journalism, improving efficiency but blurring the line between human- and machine-generated content, raising concerns about transparency and trust. While newsrooms rely on AI to meet economic and production pressures, studies show readers strongly prefer human-written content and often distrust AI involvement — especially when disclosed. This creates a paradox: transparency is demanded yet undermines confidence, leaving journalism at a critical point where maintaining credibility requires clear disclosure, human oversight, and greater public understanding of AI’s role.
An op-ed on a highly contentious issue in one of Turkey’s respected newspapers seemed sharply argued and carefully constructed — until an Artificial Intelligence (AI) prompt appeared in the middle of the text.
AI offers journalists valuable tools by processing vast volumes of data — from official documents and transcripts to videos and social media — at a speed and scale previously impossible. It can synthesize complex material and support visual storytelling, thereby expanding journalism’s analytical and narrative capacity. But how were we supposed to trust journalism that was openly — or carelessly — infused with AI?
The line between human and AI-generated journalism is blurring, and this trend is, of course, not unique to Turkey. At least 9 percent of all news articles in U.S. newspapers contain some AI-generated text, according to a 2025 University of Maryland study. It is not limited to news coverage — the same study found 219 AI-generated opinion pieces in legacy media such as The New York Times, The Wall Street Journal, and The Washington Post, most of which were written by guest contributors.
It is hardly surprising that journalists working under pressure turn to easy-to-use tools. But insufficient AI training, a poor understanding of its limitations, and the disregard for disclosing AI use can undermine journalism — an industry driven by trust — and leave readers lost in a maze of uncertainty.
Moreover, AI-generated patterns are often easy to detect — including repetitive phrasing, generic tone, and cold, formulaic sentence structure. As awareness grows, so does discomfort. Only 12 percent of readers feel comfortable with fully AI-generated news, compared to 62 percent for human-written content, according to a 2025 Reuters survey conducted across six countries.
A Hurricane Is Haunting Newsrooms — The Hurricane of AI
AI has triggered a hurricane in the newsrooms by changing many routine journalistic practices. Its integration into newsroom workflows now spans multiple stages of production — from transcribing interviews and summarizing documents to translating international reports and generating visuals.
For instance, many outlets increasingly rely on AI tools to translate international content or generate background information. The use of AI-generated visuals also became widespread because it eliminates the cost of licensing images or sending reporters into the field to capture real photographs.
AI can summarize hundreds of pages of documents or hours of video footage, saving journalists days of work. It is economical and practical — at least at first glance. The question that remains less clear is how — or whether — audiences are informed when these tools shape the content they read — and what happens when they are informed.
Despite AI’s valuable offerings, investment in training is often limited; the lack of knowledge and irresponsible use of AI in newsrooms — particularly in small, under-resourced outlets — raises concerns about long-term public trust. Proper editorial oversight and transparent disclosure are foundational pillars of modern journalism.
For readers, outsourcing arduous journalistic processes to a machine is unacceptable. Readers want to see a human touch: nearly 99% said it is important for a human to be involved in reviewing news content when AI is used, according to Trusting News’ report, AI research with LMA newsrooms’ audiences reinforces need for transparency: “Transparency is key — and people want more information, not less.”
AI Use Is Rarely Disclosed
Newsrooms widely adopt AI tools yet hesitate to acknowledge their use. A recent study found its use was rarely disclosed. A manual audit of 100 AI-flagged articles found that only five provided a citation indicating that the content contained AI.
Many newspapers today operate under intense financial pressure, paradoxically exacerbated by the emergence of AI-generated summaries in Google search results and social media algorithms. Outlets face shrinking staff and declining advertising revenue.
Even major institutions are restructuring. Recently, The Washington Post, one of the most prominent newspapers in the U.S., laid off hundreds of employees, a move its former executive editor Marty Baron said “ranks among the darkest days” in the newspaper’s history.
To stay alive in the new information age, newspapers have become laser-focused on “producing” content at a fast pace. Speed has become the new norm. And AI tools promise efficiency without additional payroll costs. Publicly acknowledging AI assistance, however, may invite skepticism or backlash from audiences already wary of automation.
Some outlets may not yet have formal, clear AI policies. Others may view AI use as a technical workflow decision rather than an editorial one. In such cases, disclosure may not be seen as necessary. Yet when AI moves into drafting or framing content, the ethical stakes shift — so do the readers’ expectations.
As Daniel Trielli, assistant professor of media and democracy at UMD’s Philip Merrill College of Journalism, told Maryland Today, “What’s most jarring is how many newsrooms are actually using this technology without saying they’re using it.”
However, when newspaper readers saw an AI use disclosure, according to Trusting News research, they almost always ask for more detail about “how” and “why” AI was used — and the degree of “human oversight” in the related article. In other words, readers are not satisfied with generic disclosures — they explicitly want total transparency.
Press enter or click to view image in full size
The Transparency Paradox
There lies a paradox. Even though readers demand greater transparency about AI use in a specific story or piece of news, such disclosures often lead to a decline in trust. Seeing AI listed as a contributor makes them uncomfortable, and they are more likely to report decreased trust than those who did not see any disclosure. Many worry that a story shaped by a “robot” could be misleading, and mere artificiality creates confusion — or unpredictability.
And so I return to the beginning: the moment I noticed an AI prompt sitting in the middle of an op-ed. In that instant, the arguments I had admired began to feel hollow. What I had taken for sharp reasoning suddenly seemed mechanical, manipulative, scary. But why?
The dystopian perception of AI, in other words, AI illiteracy, prompts fear. People want to know “how” and “why” AI was used in a specific story, but knowledge of AI’s use also undermines their trust. Transparency, instead of reassuring audiences, can activate anxieties. Trusting News found that 42% of respondents reported being “less likely” to trust the story after seeing an AI disclosure.
Therefore, newsrooms are caught in the middle. They have to choose between short-term trust by hiding AI use and long-term credibility through disclosure and honesty with their readers. At first sight, it looks like a strategic tension. However, AI disclosure is a matter of basic journalistic ethics and responsibility. Used without explanations, AI risks undermining long-term trust — and journalism itself.
“The first thing people want is education around AI,” said Lynn Walsh, assistant director of Trusting News and an Emmy Award-winning journalist. She emphasized that being caught using AI without transparency damages trust far more deeply, and that real news consumers don’t only demand disclosure but also general AI education to better understand and learn how to detect it.
More than 80% of those 6,000 survey respondents said it would be helpful if the newsroom provided information and tips to better understand AI in general, according to the Trusting News survey.
“We need to do explainers on our use of this technology, record a video as we are using it to create an image, and show them our process,” Walsh said. “That type of content might bring the audience along with us more than just having disclosures.”
Walsh underlined the importance of checking in with the audience: “Make sure that they understand that your content is still going to be accurate, responsible, and ethical.”
When AI Enters Opinion
Today, AI use in journalistic practices is unavoidable — and, in many cases, practical. But opinion writing is different. Op-eds shape public discourse, and readers expect to encounter genuine human perspectives rooted in distinct socio-economic backgrounds with different interests. They want to engage with the judgment and moral risk of a real person.
Readers, stuck in a maze of uncertainty, struggle to discern whether the opinions they consume are written by humans or generated by AI.
Using AI to test ideas or challenge one’s assumptions can be intellectually stimulating. Although AI has been criticized for reinforcing existing biases and for being a modern form of sycophancy, it can easily be guided through prompting and customization. In that sense, AI may serve as a preparatory tool for opinion writing.
But allowing AI to write an opinion piece is analogous to publishing an AI-generated poem as if it were written by a human. It risks hollowing out authentic human exchange and weakening the very deliberation that opinion writing is meant to foster.
AI-generated content in opinion pages across legacy media accounted for 4.5% overall — significantly higher than the rate on those outlets’ news pages (0.7%), according to the University of Maryland study.
When the human presence is replaced — or concealed — on the opinion pages, the relationship between writer and reader erodes. Delegating an opinion essay to AI without disclosure undermines public trust in newspapers.
Independent AI-Detection Tools
When readers suspect articles are generated by AI without disclosure, credibility erodes. In journalism, diminished trust means drowning in the ocean of information, as lower engagement and declining advertising revenue hinder quality journalism. If journalism depends on trust as its economic and moral foundation, embedding multi-dimensional transparency around AI use is essential.
Industry-wide standards must be established, and disclosure must become a baseline requirement. News organizations should adopt and publicly make available internal policies specifying when AI may be used and when it may not.
Any meaningful AI involvement in reporting, editing, or visual generation should be clearly labeled. Readers are more likely to accept AI assistance when they understand how and why it was used. It is a process that should advance in communication with the readers, constantly engaging with their expectations and discomforts.
Moreover, readers should be able to access independent, credible AI-detection tools — such as systems modeled on Pangram — specifically for journalistic contexts. They would enable readers to verify the origins of content, strengthening accountability and rebuilding confidence in the press.
By combining transparency, AI literacy, and independent detection tools, journalism can protect its foundational principle: trust.
“Journalists make more informed decisions about what it means to serve the public if they understand the people that they are trying to serve,” Walsh said.