Can AI Save Local Journalism Without Weakening It?

Across African newsrooms and media houses, artificial intelligence is reducing the time spent on tasks like transcription, research and digital production. Whether that relief produces stronger local journalism or substandard work bereft of the qualities of a good story will depend on what newsrooms refuse to hand over.

Years ago, when Lekan Otufodunrin worked as a Punch correspondent in Abeokuta, Ogun State, filing a story required two people and considerable patience. Once his report was ready, he called the newspaper’s head office in Lagos and dictated every word to someone at the other end. The colleague wrote it down before the story was typed again for publication.

Computers eventually displaced handwriting, and email reduced the need for telephone dictation. With the advent of technology, mobile phones have placed cameras, recorders and publishing tools in journalists’ pockets.

Otufodunrin, now Executive Director of the Media Career Development Network and Managing Editor of Nigeria Media Monitor, does not romanticise the inconvenience of that period. Journalism has always adjusted to new tools, he says, and artificial intelligence is another stage in that progression. Still, this transition feels different and more like a disruption.

Earlier tools only changed how journalists recorded, transmitted and published their work. Generative AI can participate in producing the work itself. It can propose a headline, suggest an angle, summarise evidence and produce a finished-looking report, sometimes without the journalist doing enough reporting to justify it.

This power has entered newsrooms at a vulnerable moment. Local media organisations are under pressure to publish faster, serve audiences across several platforms and defend their credibility against manipulated content. And many are trying to meet those demands with small teams and limited resources.

For these newsrooms, AI promises relief. It also raises a harder question: what happens when the technology taking work off journalists’ hands begins taking over the thinking that makes the work journalism?

Africa’s AI Moment: Are We Building or Just Buying? 

Why overstretched newsrooms cannot ignore AI

The local reporter’s job has expanded beyond finding and filing a story. The same reporting may now have to become a website article, television package, newsletter item and social media video, often without more people being hired to produce it.

This workload has made the case for AI unusually persuasive. If the technology can shorten one stage of production, it gives an overstretched newsroom something it rarely possesses: more time.

Matthew Ogunsuada understands how quickly that time disappears. As a broadcast journalist with Voice of the People TV, his working day is shaped by events that rarely wait for a convenient production schedule.

A fire breaks out, a road accident brings traffic to a standstill, or a political announcement changes an earlier story. The reporter has to move quickly because broadcast news loses value as events overtake it.

For Ogunsuada, he has to be on his toes to research the subject, monitor sources, verify claims, conduct interviews, write the script, shoot and then edit the video for dissemination. Afterwards, he may still need to prepare clips and teasers for digital audiences because, well, the younger generation is chronically online and he can’t afford to lose them.

Speaking on how the broadcast journalist manages to do all these in the age of AI, he said, “When I am overwhelmed, maybe I am working on a story, and it is becoming too long, I first write what I want or draft an idea of what I want to do, then I bring AI in to help with the transcriptions, summarising of supporting documents and refining of script.”

However, while tools such as Otter and TurboScribe help with transcription, and generative AI helps refine a script he has already prepared, the main idea and final editorial decision remain his. 

Further research: a 2025 study by the Center for News, Technology & Innovation surveyed 109 journalists across Africa, including 75 in Nigeria. It found that respondents were more optimistic about technology’s impact on journalism than journalists elsewhere and more likely to use technological tools for document analysis, summarisation and writing support. 

The attraction is clear and not far-fetched.  Sorting a government document quickly can help uncover a lead. Saving two hours on transcription can change what a reporter completes that day. Translating a report into multiple languages can open it to an audience; the publication lacks the staff to serve manually.

Now that AI has arrived as an extra pair of hands, what type of work should newsrooms allow those hands to perform? Which should remain the full responsibilities of the learned and professionals?

The tasks AI is already taking over

Public fears about AI in journalism often centre on machines generating complete articles with zero human input. In practice, its most useful applications sit around the reporting process, handling repetitive or cumbersome tasks like research and transcriptions without necessarily determining the story.

Peace Chenube, an investigative journalist at the Foundation for Investigative Journalism, primarily uses generative AI for research and comprehension. Some documents he examines run into tens of pages. AI helps him search public records and identify passages deserving closer attention.

The output does not automatically become evidence because a chatbot discovered it. Chenube mentions returning to the source document and establishing what the information means within the investigation. 

Similarly, a journalist at one of Africa’s leading technology news outlets who requested anonymity uses AI when an assignment takes her into an unfamiliar subject. One story may involve cryptocurrency, another blockchain infrastructure and the next an unfamiliar tax rule.

To understand any of these industries, she asks the system to explain the concept as though she were five years old, using familiar objects to create an illustration she can understand. The explanation gives her a starting point for further research.

AI may also suggest angles, but she distrusts its handling of names, figures and quotations. When she asks it to extract quotes from a transcript, she requests time codes and checks them against the original recording.

“Accuracy is all on me,” she said. “Everything the agent gives back to me, I verify. If it is not accurate, I drop it.”

Larger news organisations are developing more advanced versions of these workflows. The Philadelphia Inquirer built Scribe to monitor public meetings, convert transcripts into summaries and rank possible developments using a framework created by journalists and editors. The Associated Press has also used AI-supported tools to examine court filings and assist with visual verification.

These examples suggest that AI’s strongest contribution may be helping reporters find stories hidden inside repetitive work. However, examples published by OpenAI represent the perspective of an interested technology provider. They demonstrate what participating organisations are building, not whether those projects can be afforded or sustained by smaller newsrooms.

AI is also changing the infrastructure through which journalism reaches its audience. This is evidenced in the work Ajiri Omafokpe is doing across the technology ecosystems of Africa and the Gulf. The Qatar-based AI builder and co-founder of BuildersTribe has used AI-assisted coding to build multiple websites in recent times.

To her, these tools drastically reduced the technical barrier, but generating code is only the beginning. She still had to define the platform’s requirements and repeatedly test its user journeys to ensure readers get value and have the best user experience.

One of the platforms she built includes French and Arabic functionality for readers across Africa and the Gulf Cooperation Council, alongside accessibility considerations for visually impaired users. That way, she can cater to a more diverse audience.

These applications and developments could bring more community voices into journalism. They also make it easier to generate finished-looking content, leaving editors to determine how much original thought and reporting sits underneath it.

Faster stories, weaker journalists?

A polished draft can create the impression that the difficult part of journalism has been completed. It, however, takes editorial judgment to recognise when fluent writing conceals a weak angle, inadequate sourcing or the absence of original thought.

That problem becomes more serious when cub reporters learn automation before developing the competence needed to evaluate its output.

Daniel Nnamani encounters this as Editor-in-Chief of UNILAG Sun, the community newspaper serving the University of Lagos. The publication documents policies, events and people across the university while giving young reporters practical newsroom experience.

Nnamani recounted one recent submission that tested his patience. It was peak season for the newsroom, and the pressure to turn in deliverables at the speed of light was high. So he assigned writers to cover a beat.

When it was submitted, the headline was poor, the lead failed to present the news clearly, and traces of an AI-generated response remained inside the copy. He was so furious at the culprits and lamented the negative effect AI is having on student journalists. 

“AI is becoming a huge challenge in training journalists because a lot of them, especially student journalists who might not have plans to pursue a career in journalism, just want to get it done quickly. They put it on AI and deliver whatever results they get. They don't cross-check or edit. But that's not to say that AI is bad when you are writing. You can incorporate it, but do so reasonably,” he said.

Nnamani may use AI to improve a headline he has already developed or identify problems in a completed draft. In those cases, he brings an editorial decision to the tool. The reporters he worries about expect the tool to make that decision for them.

In another expert opinion, Professor Ibitayo Popoola, Head of the Department of Mass Communication at the University of Lagos and an editorial adviser to UNILAG Sun, said he has seen AI increase the speed at which students produce stories. And as Nnamani identified earlier, some of the reports lack depth and authenticity.

“Journalism is brain work,” Popoola said. “On no account must you allow AI to replace the traditional way of thinking and acting journalistically. Overreliance on AI will diminish your ability to create. Journalism is a job of creativity.

"Creativity in writing, a stylistic approach, and the innovative way you package your story. Transferring this responsibility to AI is bad. It will reduce the nuances and contextualisation of stories, which in turn will affect the authenticity of media and news reports.”

At the risk of sounding archaic, the Professor welcomes tools that remove manual transcription, provided journalists still check names and unfamiliar expressions. His concern begins when students outsource the creative and intellectual stages of reporting.

Otufodunrin also sees the consequences among graduates who understand AI tools but struggle with interviewing, observation and basic news writing. “AI can help you rewrite and tune up your report, but it will not help you to see the story that you are not seeing,” he said. 

From all these takes and expert opinions, it is obvious that the technology is most useful after a journalist has recognised a problem and gathered relevant material. When it arrives before those decisions, it gives a young reporter another opportunity to avoid developing judgment.

The context the machine cannot see

At its best, generative AI is still the average of all the smartest human intelligence fed into it over time and can be confidently wrong in its output. This situation is generally known as AI hallucinations.

According to IBM, AI hallucinations are instances where an AI system produces outputs that sound plausible but are factually wrong, irrelevant or entirely fabricated.

In these instances, a large language model (LLM) or other tool presents fake facts, invented studies, nonexistent URLs or incorrect details about real entities as factual outputs.

Because fluency can make incomplete information appear authoritative, a journalist must know enough about the subject to recognise what has been distorted or omitted. This becomes particularly consequential when AI interprets communities that are poorly represented online.

Nnamani shared a real-life example. While researching the light situation in Makoko, he wanted to know the electricity tariff band covering the Lagos community. Instead of answering the question, the AI tool produced a broad and false description of Makoko as a place without electricity, toilets or public infrastructure.

Although Makoko contains neighbourhoods with better structures and living conditions, the system reconstructed the version most visible in public material, reducing the community to its most familiar image.

This failure reveals whose accounts shape how AI interprets African places. Communities repeatedly reported through poverty, conflict, or disease may reappear through the same narrow frames. The information could sound plausible while sustaining a misleading picture, but one of the hallmarks of journalism is accuracy. And that is non-negotiable.

When the verification tool becomes the threat

AI is entering newsrooms as a research assistant while making the external information environment harder to trust. The same technology helping journalists examine evidence can produce convincing records of events that never occurred.

In December 2025, Chenube encountered a video that appeared to show foreign soldiers storming a beach in Nigeria. It circulated while public concern about insurgency and terrorism was already heightened.

The investigative journalist tried to trace its origin and establish whether the incident had occurred. He could not confirm the video’s authenticity and suspected it was AI-generated.

The footage showed how synthetic media can exploit an existing public fear. Its viewers may not demand strong evidence when the images resemble a threat they already consider possible.

Because of a strict verification-before-reporting rule at FIJ, whenever Chenube cannot ascertain the credibility of a photograph or video directly, he uses tools such as Google Reverse Image Search and Yandex to investigate its origin. 

The technology journalist who decided to stay anonymous also applies a similar suspicion to less dramatic outputs. Every name, number and quotation generated or extracted by AI returns her to a source.

This scrutiny requires time, and Ogunsuada’s experience shows the difficulty: journalists are expected to authenticate increasingly sophisticated material without losing the speed demanded by broadcast and digital news.

But for Popoola, accuracy remains immovable. “If a medium loses its credibility, that is its end as a medium of information dissemination,” he said.

Who can afford responsible AI?

Responsible adoption requires reliable tools, staff training, editorial supervision, privacy protections and clear rules governing what information can be submitted to external systems.

Those safeguards cost money. Their uneven availability may determine which newsrooms benefit from AI and which inherit its greatest risks.

Nnamani describes AI as both a survival tool and a luxury. It helps during busy production periods when a small editorial team has several reports to review. The luxury lies in the recurring subscriptions required for more capable versions of popular products. 

A larger publisher may acquire enterprise accounts, develop internal tools and establish data-protection policies. Smaller outlets may have journalists experimenting through personal accounts without common standards or clarity about how uploaded material will be stored.

Institutional policies are still catching up with individual use. Broadcast Media Africa’s 2026 readiness reporting found that AI was expanding across media production while many organisations lacked sufficient safeguards for accuracy, transparency and accountability. Its reporting also suggested that infrastructure and technical-capacity gaps continued to limit AI’s commercial impact. 

The disparity becomes clearer when investments elsewhere are considered. In Qatar, the Government Communications Office partnered with the Google News Initiative and Al Jazeera Media Institute to train journalists in AI and digital transformation. Al Jazeera later announced “The Core,” an AI-integrated operating model developed with Google Cloud. 

Many small African publications are operating far from this level of investment. For some, an AI strategy may consist of one reporter’s personal subscription and whatever guidance an editor can offer.

To this end, Otufodunrin believes journalism schools must help close the capacity gap by incorporating AI into their curricula. But the lecturers and instructors need training too. A revised course will accomplish little if those teaching it do not understand the tools already moving through their students’ work. 

Funding relationships raise another concern. OpenAI’s partnership with Axios finances local newsletters while allowing OpenAI to train its models on Axios’s published reporting. The deal tests whether AI-assisted local operations can become commercially viable, but it also gives a technology company access to the information its systems require. 

Jennifer Brandel argues in Nieman Lab that AI companies may increasingly support local journalism because their models need current, verified information. As synthetic material spreads across the web, original reporting becomes essential infrastructure for maintaining the quality of AI-generated answers. 

If African reporting supplies knowledge that cannot be gathered elsewhere, publishers and AI builders should recognise its value. Access to software is not necessarily fair compensation for the reporting on which the software depends.

What must remain human for local journalism?

The precise boundary around newsroom AI will vary by assignment, house style or organisation. An investigative reporter handling leaks cannot operate by the same rules as a desk adapting a public announcement for social media.

Yet the experts cited in this report reveal a chain of responsibility worth protecting. Observation produces the idea. Sources supply knowledge. Reporting establishes context. Verification tests the evidence. Editorial judgment determines what can be published.

Chenube draws a boundary around ideation, verification and drafting. An investigation may begin when a reporter notices that a bus arrives late at the same stop every morning and starts asking whether the repetition points to a wider failure. The story exists because somebody was present long enough to recognise the pattern.

Sources also enter through human relationships. People decide whether to share an experience or trust a reporter with information that could place them at risk. AI cannot build that trust or accept responsibility if the resulting evidence is misunderstood.

Also for Chenube, confidential records establish a firm boundary. He restricts his AI use to publicly available information. Documents containing source identities or other sensitive details remain outside commercial systems. 

Other interviewees draw their lines differently. Ogunsuada uses AI to refine a script. Nnamani may use it to improve a headline. The technology reporter consults it for explanations while independently verifying every factual output.

Their practices differ, but an identifiable person remains responsible at each decisive point. Omafokpe reaches a similar conclusion from the product side. AI can lower the cost of building a publication, but someone must still define its purpose, understand its audience and determine the ethical boundaries governing its content.

Otufodunrin believes experienced journalists possess an advantage if they are willing to learn again. They understand the work manually and can make informed decisions about where technology improves it. His words: “Get enabled. Get the skills and be ready to fit into the new reality.” 

Journalism does not preserve its standards by refusing new tools. It preserves them by ensuring that reporters know which responsibilities cannot be transferred.

For newsrooms with skilled reporters, clear policies and strong editorial supervision, AI can free up time for deeper work. Where those foundations are missing, it can conceal weak journalism long enough for it to reach the public.

Years ago, the telephone carried Otufodunrin’s reports from Abeokuta to Lagos, but it never decided what the story was. Generative AI has blurred that boundary.

Learning to use it is becoming part of working in a modern newsroom. Learning when not to use it may prove equally important.

Before AI can help a journalist tell a story, there must still be a journalist capable of finding one. And before AI can help save a local newsroom, it must be grounded in all the hallmarks of true journalism.